Kfkgig Fifd 4u

Transformer Winding Deformation and Insulation
Characteristic Effects on Frequency Response Analysis
Author:
Bagheri, Mehdi
Publication Date:
2014
DOI:
https://doi.org/10.26190/unsworks/16831
License:
https://creativecommons.org/licenses/by-nc-nd/3.0/au/
Link to license to see what you are allowed to do with this resource.
Downloaded from http://hdl.handle.net/1959.4/53515 in https://

unsworks.unsw.edu.au on 2023-12-04
Transformer Winding Deformation and
Insulation Characteristic Effects on
Frequency Response Analysis
By
Mehdi BAGHERI
Supervisor: Dr.Toan PHUNG

Co-supervisor: A/Prof. Trevor BLACKBURN
A THESIS IN FULFILMENT OF THE REQUIREMENT FOR

THE DEGREE OF DOCTOR OF PHILOSOPHY
SCHOOL OF ELECTRICAL ENGINEERING AND TELECOMMUNICATIONS

FACULTY OF ENGINEERING
March 2014
THE UNIVERSITY OF NEW SOUTH WALES
Thesis/Dissertation Sheet
Surname or Family name: BAGHERI
First name: Mehdi Other name/s:
Abbreviation for degree as given in the University calendar: Ph.D.
School: Electrical Engineering and Telecommunications Faculty: Engineering
Title: Transformer Winding Deformation and Insulation Characteristic Effects on Frequency Response Analysis
Abstract 350 words maximum:

Frequency Response Analysis (FRA) is considered an accurate, fast, economical and non-destructive method for
the detection of winding deformation within power transformers, providing detailed information on electrical
properties of this asset. Changes in winding configuration, as well as other transformer active part structures
would almost certainly cause variation in the frequency response spectrum. This can be exploited for mechanical
defect recognition. On the other hand, transformer oil deterioration, temperature variation as well as water
absorbed by the paper can cause transformer insulation characteristics to change over the time. In fact,
capacitances, self- and mutual inductances and conductor resistances might be altered due to any changes in
above mentioned factors. In turn, the frequency response of the winding will change accordingly. Thus in the
interpretation of the FRA spectrum for evidence of winding deformation, the influence of insulation
characteristic on the spectrum must be taken into consideration. FRA deviation due to the winding deformation
or insulation characteristic changes becomes even more complicated to interpret when FRA baseline and
measured spectra are taken under different temperatures and moisture contents. In such a case, existing FRA
evaluation methods using statistical indicators are likely to reveal incorrect prognosis. Hence in this thesis, the
aim of the research is to distinguish the insulation characteristic impacts on FRA spectrum from winding
deformation. To this end, resonances and anti-resonances in FRA spectrum over different frequency bands are
examined in detail and interpretations are provided. FRA deviation due to the transformer winding deformation
is discussed analytically, modelled and simulated. The results are then compared to practical measurements.
Insulation characteristic changes in transformer are studied through temperature and moisture variations to
recognise their influences on FRA data. FRA capability in recognising moisture migration from the paper
insulation of transformer winding is recommended and its potential application in transformer winding dry-out
process evaluation is revealed in this research. Finally, possible offline and online solutions to distinguish the
impact of moisture and temperature variations on winding deformation diagnosis are provided. Online FRA
measurement and its required setup as a potential future approach in transformer condition monitoring are
discussed.
Declaration relating to disposition of project thesis/dissertation
I hereby grant to the University of New South Wales or its agents the right to archive and to make available my thesis or
dissertation in whole or in part in the University libraries in all forms of media, now or here after known, subject to the
provisions of the Copyright Act 1968. I retain all property rights, such as patent rights. I also retain the right to use in future
works (such as articles or books) all or part of this thesis or dissertation.
I also authorise University Microfilms to use the 350 word abstract of my thesis in Dissertation Abstracts International (this is
applicable to doctoral theses only).
……………………………………………………… …………………………………………… ……14/06/2014……….

Signature Witness Date
The University recognises that there may be exceptional circumstances requiring restrictions on copying or conditions on use.
Requests for restriction for a period of up to 2 years must be made in writing. Requests for a longer period of restriction may
be considered in exceptional circumstances and require the approval of the Dean of Graduate Research.
FOR OFFICE USE ONLY Date of completion of requirements for Award:

COPYRIGHT STATEMENT
‘I hereby grant the University of New South Wales or its agents the right to
archive and to make available my thesis or dissertation in whole or part in
the University libraries in all forms of media, now or here after known,
subject to the provisions of the Copyright Act 1968. I retain all proprietary
rights, such as patent rights. I also retain the right to use in future works
(such as articles or books) all or part of this thesis or dissertation.
I also authorise University Microfilms to use the 350 word abstract of my
thesis in Dissertation Abstract International (this is applicable to doctoral
theses only). I have either used no substantial portions of copyright material
in my thesis or I have obtained permission to use copyright material; where
permission has not been granted I have applied/will apply for a partial
restriction of the digital copy of my thesis or dissertation.'
Signed…………………………………………..............
Date……………14/06/2014.…………………..............
AUTHENTICITY STATEMENT
‘I certify that the Library deposit digital copy is a direct equivalent of the final
officially approved version of my thesis. No emendation of content has
occurred and if there are any minor variations in formatting, they are the
result of the conversion to digital format.’
Signed…………………………………………..............
Date……………14/06/2014.…………………..............
ORIGINALITY STATEMENT
‘I hereby declare that this submission is my own work and to the best of my
knowledge it contains no materials previously published or written by
another person, or substantial proportions of material which have been
accepted for the award of any other degree or diploma at UNSW or any
other educational institution, except where due acknowledgement is made
in the thesis. Any contribution made to the research by others, with whom I
have worked at UNSW or elsewhere, is explicitly acknowledged in the
thesis. I also declare that the intellectual content of this thesis is the product
of my own work, except to the extent that assistance from others in the
project's design and conception or in style, presentation and linguistic
expression is acknowledged.’
Signed…………………………………………..............
Date……………14/06/2014.…………………..............
A THESIS DEDICATED TO MY FATHER AND SISTER
IN LOVING MEMORY OF MY MOTHER

Acknowledgment
I would like to express my sincere gratitude to my supervisor, Dr. Toan PHUNG, for his
continuous support throughout my PhD study. I cannot thank enough for his guidance,
insightful instruction, and ceaseless encouragements during the past years.
I would also like to give my special appreciation to my co-supervisor, A/Prof. Trevor

BLACKBURN. His inspiring minds and firm support have definitely taken my research to
another level. I have been deeply influenced by his enthusiasm for science, ambitious heart
in research, and greatest dedication to work.
In addition, I would deeply thank Dr. Mohammad SALAY NADERI my former supervisor
for his detailed instruction and fruitful discussions. His ongoing and valuable advice
allowed me to expand my knowledge broadly and gain very useful skills.
Last but not least, I am forever indebted to my father and sister, who gave me a beautiful
life and supported me for studying overseas emotionally. I also give my special thank to
my mother who could not survive to see this great time in my life. It is your love that
embraced me through the PhD and my life.
Abstract
Frequency Response Analysis (FRA) is considered an accurate, fast, economical and non-
destructive method for the detection of winding deformation within power transformers,
providing detailed information on electrical properties of this asset. Changes in winding
configuration, as well as other transformer active part structures would almost certainly
cause variation in the frequency response spectrum. This can be exploited for mechanical
defect recognition. On the other hand, transformer oil deterioration, temperature variation
as well as water absorbed by the paper can cause transformer insulation characteristics to
change over the time. In fact, capacitances, self- and mutual inductances and conductor
resistances might be altered due to any changes in above mentioned factors. In turn, the
frequency response of the winding will change accordingly. Thus in the interpretation of the
FRA spectrum for evidence of winding deformation, the influence of insulation characteristic
on the spectrum must be taken into consideration. FRA deviation due to the winding
deformation or insulation characteristic changes becomes even more complicated to
interpret when FRA baseline and measured spectra are taken under different temperatures
and moisture contents. In such a case, existing FRA evaluation methods using statistical
indicators are likely to reveal incorrect prognosis. Hence in this thesis, the aim of the
research is to distinguish the insulation characteristic impacts on FRA spectrum from
winding deformation. To this end, resonances and anti-resonances in FRA spectrum over
different frequency bands are examined in detail and interpretations are provided. FRA
deviation due to the transformer winding deformation is discussed analytically, modelled
and simulated. The results are then compared to practical measurements. Insulation
characteristic changes in transformer are studied through temperature and moisture
variations to recognise their influences on FRA data. FRA capability in recognising
moisture migration from the paper insulation of transformer winding is recommended
and its potential application in transformer winding dry-out process evaluation is revealed
in this research. Finally, possible off-line and on-line solutions to distinguish the impact of
moisture and temperature variations on winding deformation diagnosis are provided.
Online FRA measurement and its required circuit setup as a potential future approach in
transformer condition monitoring are discussed.
i
Nomenclature
A Sinusoidal signal amplitude Ch.2

Ac Cross sectional area of core limb Ch.5
Arog Cross sectional area of each small loop in Rogowsky coil Ch.8
Ay Cross sectional area of core yoke Ch.5
B Magnetizing flux density Ch.2, Ch.5
c Local moisture concentration Ch.7
cs Series capacitance per unit length Ch.4
cg Shunt capacitance to ground per unit length Ch.4
C Total winding capacitance Ch.5
Cc Substance concentration Ch.7
Cd Total series capacitance of entire disks Ch.3
Ci Capacitance at ith resonance frequency Ch.7
Cs Total series capacitance Ch.3
Ct Total turns’ capacitance Ch.3
Ctt Turn-to-turn capacitance Ch.3
Cs-pair Pair-disk equivalent series capacitance Ch.3
CgHV(meas) Total measured shunt capacitance of winding Ch.3
CgHV(calc) Total calculated shunt capacitance of winding Ch.3
Csh(b) Winding shunt capacitance of phase b when phase B isolated from
Ch.5
the ground
Cʹsh(b) Winding shunt capacitance of phase b when phase B is grounded Ch.5
CHT Shunt capacitance of HV winding respect to transformer tank Ch.5
CHL Shunt capacitance of HV winding respect to the LV winding Ch.5
CLC Shunt capacitance of LV winding respect to the core Ch.5
Cdeform Shunt capacitance of deformed section Ch.6
Cnorm Shunt capacitance of normal section Ch.6
C0 Absolute permittivity of vacuum Ch.7
Cʹg Shunt capacitance of buckled winding Ch.6
Cʹtt Total turn-to-turn capacitance of the deformed disk Ch.6
d Distance between circular turns Ch.3
dʺ Average length of conductor turn Ch.4
dAc Infinitesimal of the core cross section area Ch.5
dp Paper insulation thickness or pressboard Ch.7
D Diffusion coefficient Ch.7
Dave Winding average diameter Ch.2
D0 Pre-exponential factor Ch.7
Dielectric displacement Ch.7
ii
Eint Number of disks used for interleaving Ch.3
e Potential to ground (time domain) Ch.4
ē Induced voltage Ch.5
er Measurement error Ch.8
E0 Activation energy of diffusion process Ch.7
Electric field Ch.7
Ed Total disk-to-disk energy App. C
Et Total turn-to-turn energy App. C
Etot Total energy stored in a pair of disks App. C
f Operational frequency Ch.2
fi ith resonance frequency Ch.7
fr Resonance frequency Ch.4
fr-low Anti-resonance frequency Ch.5
fsweep Variable frequency Ch.2
Fradial Imposed force during transportation Ch.2
Fpredicted Predictable force Ch.2
Fstop Stop force Ch.2
Fwind Wind force Ch.2
g Shunt turn-to-turn conductance per unit length Ch.4
G Shunt conductance to ground Ch.4
h Axial dimension of conductor cross section Ch.3
Hm1 Height of LV winding Ch.2
Hm2 Height of HV winding Ch.2
Hw Winding height Ch.2, Ch.3,
and Ch.6
i Conductor current Ch.4, Ch.5
icg External capacitive current per unit conductor length Ch.4
ics Internal capacitive current per unit conductor length Ch.4
ig External conductance current per unit conductor length Ch.4
iG External conductance current per unit conductor length Ch.4
i2100 Current square harmonic at 100 Hz Ch.8
I Filament current App. B
I(jω) Conductor current (frequency domain) Ch.4
Ish Short-circuit current Ch.2
Iinsul Current traversing the medium Ch.7
j Imaginary operator Ch.4
J The current density vector App. B
ka LV winding frequency response magnitude of phase a Ch.5
kaB(sc) LV winding frequency response magnitude of phase a when phase B
Ch.5
of HV side is short-circuited
kaC(sc) LV winding frequency response magnitude of phase a when phase C
Ch.5
of HV side is short-circuited
kʹ Dimensionless parameter and is equal to 0.5 Ch.7
kʹa Generated frequency response magnitude of phase a Ch.5
k0 Circular inductance decrement factor Ch.3
iii
Km Mass coefficient of oil App. F
Kmag Frequency response magnitude Ch.2
KN Nagaoka’s coefficient Ch.3
KR Rogowsky coefficient Ch.2
l Inductance per unit length Ch.4
lc Mean magnetic path length of core limb Ch.5
ly Mean magnetic path length of yoke Ch.5
Lʹ The self-inductance of a single non-circular filament App. B
Leq Winding equivalent inductance Ch.3
LHV(calc) Total calculated inductance of winding Ch.3
LHV(meas) Total measured inductance of winding Ch.3
L Total winding self-inductance Ch.5
La LV winding inductance of phase a Ch.5
LaB(sc) LV winding inductance of phase a when phase B of HV side is short-
Ch.5
circuited
LaC(sc) LV winding inductance of phase a when phase C of HV side is short-
Ch.5
circuited
Li Inductance at ith resonance frequency Ch.7
LOC HV winding inductance at 50 Hz when LV side terminals are left
Ch.6
open circuit
LSC HV winding inductance at 50 Hz when LV side terminals are short-
Ch.6
circuited
L1 Self-inductance of primary winding Ch.8
L2 Self-inductance of secondary winding Ch.8
m Total transformer mass Ch.2
M Mutual-inductance of coaxial circular filaments Ch. 3, Ch.6
Mʹab The mutual-inductance between a circular and non-circular
App. B
filaments
Mʺab The mutual-inductance between two non-circular filaments App. B
Mc Disk-to-disk mutual-inductance Ch.3
Meq Winding equivalent mutual-inductance Ch.3
Mʹ Mutual-inductance of the circular filaments whose axes inclined to
Ch. 6
one another
Mʹʹ Mutual-inductance between the turns for asymmetrical axial
Ch.6
deformation of a disk
Mʹʹʹ Mutual-inductance of circular elements with parallel axes Ch.6
n Turn number Ch.3
N Number of disk turns Ch.3
NB Number of transformer core limbs surrounded by HV and LV
Ch.2
windings
Nd Number of transformer winding disks Ch.3
Nt Number of winding turns Ch.2
NI R.M.S. winding’s ampere-turns value Ch.2
Nw Number of winding turns Ch.3, Ch.4,
and Ch.5
Nsh Number of shield turns per disk Ch.3, App.C
iv
Ns Number of elements (samples) App. A
Nrog Number of coil turns in Rogowsky coil Ch.8
P Total mass of paper scraps App. F
q Coupling coefficient Ch.8
r Conductor resistance per unit conductor length Ch.4
r The distance between the filament center and associated point for
App. B
flux density
r' Radius of deformed section Ch.6
r1 Average radial dimension of winding Ch.3
r2 Average radial dimension of tank Ch.3
R Mean radius of disk Ch.3
Ra Mean radius of circular turn a Ch.3, Ch.6,
and App.B
Rb Mean radius of circular turn b Ch.3, Ch.6,
and App.B
Rd The distance between the current source and the associated point
App. B
for flux density
Rn Resonance ratio number Ch.7, App. G
S Transformer apparent power Ch.2
Si Measured scatter parameters Ch.8
Sref Scatter parameter Ch.8
T0 Reference temperature Ch.7
Tk Current temperature Ch.7
u Estimated uncertainty Ch.8
u2100 Voltage square harmonic at 100 Hz Ch.8
U Voltage across a pair-disk Ch.3, App. C
Uinsul Voltage across the insulation medium Ch.7
Ux Short circuit reactance Ch.2
v The volume containing current App. B
υtank-100Hz Tank vibration frequency at 100 Hz Ch.8
V(jω) Potential to ground (frequency domain) Ch.4
Vn Nominal voltage of winding Ch.2
Vin Injected voltage at line-lead Ch.2
Vout Measured voltage at neutral-lead Ch.2
VO Output voltage Ch.8
Vphase Phase voltage Ch.8
Vtap Bushing tap voltage Ch.8
Vx Shield turn potential App. C
W Radial dimension of winding cross section Ch.3
Wspacer Spacer width Ch.2
Wref Reference paper humidity in percent Ch.7
Wll Lower limit of the paper humidity in percent Ch.7
Wul Upper limit of the paper humidity in percent Ch.7
x Substance movement position Ch.7
X Axial displacement of the outermost turn Ch.6
v
Xi ith elements of the reference FRA spectrum App. A
Y Total mass of accumulated water App. F
Yi ith elements of the measured FRA spectrum App. A
Za Transformer LV winding impedance of phase a Ch.5
ZaB(sc) LV winding impedance of phase a when phase B of HV side is short-
Ch.5
circuited
ZaC(sc) LV winding impedance of phase a when phase C of HV side is short-
Ch.5
circuited
Zin Input impedance of measurement cable Ch.5
Zout Output impedance of measurement cable Ch.5
ZOC(50 Hz) Inductive reactance of HV winding at 50 Hz when LV side terminals
Ch.6
are left open circuit
Zp Impedance paralleled with bushing tap Ch.8
Zsc Short circuit impedance Ch.8
ZSC(50 Hz) Inductive reactance of HV winding at 50 Hz when LV winding is
Ch.6
short-circuited
Zw Transformer winding impedance Ch.5
Z11 Open circuit input impedance Ch.8
Z12 Open circuit reverse transfer impedance Ch.8
Z21 Open circuit forward transfer impedance Ch.8
Z22 Open circuit output impedance Ch.8
α Initial impulse voltage distribution coefficient Ch.4
αʹ The angle between Ra and Rd App. B
γ Self-inductance per unit conductor length Ch.4
δt Inter-turn insulation thickness Ch.3
ε Dielectric permittivity Ch.3
εoil Oil insulation permittivity Ch.7
εpaper Paper insulation permittivity Ch.7
εspacer Spacer permittivity Ch.7
εr Relative permittivity of dielectric Ch.7
εt Paper insulation permittivity Ch.3
ε0 Vacuum permittivity Ch.3
έ0 Static dielectric constant Ch.7
ε∞ Infinite-frequency dielectric constant Ch.7
ε' Real part of complex permittivity Ch.7
εʹʹ Imaginary part (loss factor) of complex permittivity Ch.7
ε* Complex permittivity Ch.7
η Ratio of entire trigonometric circular span Ch.6
θʹ The angle between the differential current vector and the vector
App. B
directed from it to the point associated for flux density
θs Angle between spacers Ch.2
θto Top oil temperature Ch.8
ϑa Phase angle in the first loop for mutual-inductance calculation App. B
ϑb Phase angle in the second loop for mutual-inductance calculation App. B
λ Winding length Ch.2, Ch.4
vi
λrog Rogowsky winding length Ch.8
λ' Entire conductor length in one disk Ch.3
μ Mutual inductance per unit conductor length Ch.4
μc Core permeability Ch.5
µs Static friction constant Ch.2
μ0 Vacuum permeability Ch.2, Ch.5
σ Dielectric conductivity Ch.7
σoil Oil insulation conductivity Ch.7
σpaper Paper insulation conductivity Ch.7
τ Diffusion time constant Ch.7
ϕ Magnetizing flux Ch.5
ϕb The induced magnetizing flux on the second loop due to the current
App. B
initiated by the first loop
χ Electric susceptibility Ch.7
ω Angular frequency Ch.4, Ch.8
R Magnetic core reluctance Ch.5
Ra Equivalent magnetic reluctance for phases a/A Ch.5
Rʹa Generated magnetic reluctance for phase a Ch.5
RaB(sc) Equivalent magnetic reluctance of phases a/A when phase B of HV
Ch.5
side is short-circuited
RaC(sc) Equivalent magnetic reluctance of phases a/A when phase C of HV
Ch.5
side is short-circuited
Rl Leakage reluctance Ch.5
Rc Transformer core limb reluctance Ch.5
Ry Transformer core yoke reluctance Ch.5
FSD FRA Spectrum Deviation App. G
MAMD Mean absolute magnitude distance Ch.8
MAPD Mean absolute phase distance Ch.8
RH Air relative humidity in percent Ch.7
WCO Water content in oil Ch.7
WCP Water content in paper Ch.7
WCP Moisture in paper in percent by weight Ch.7
vii
Contents
Abstract i
Nomenclature ii
Contents viii
List of Figures xiii
List of Tables xix
Chapter 1 Introduction 1
1.1 Problem Statement 1
1.2 Motivation 2
1.3 Current State of the Art 3
1.4 Research Objectives 6
1.5 Thesis Overview 7
1.6 Key Contributions 9
1.7 Publications 10
Chapter 2 Transformer Winding Deformation and Diagnosis Techniques 13
2.1 Introduction 13
2.2 Transformer Winding Deformation and Displacement 13
2.2.1 Short Circuit Current 14
2.2.2 Transformer Transportation Causing Active Part Displacement 18
2.3 Winding Deformation Diagnosis Methods 21
2.3.1 Short Circuit Impedance Measurement 21
2.3.2 Transfer Function Methods (FRA/LVI) 24
2.3.3 Deformation Coefficient Method 32
2.4 FRA vs. SCI 33
2.4.1 SCI Measurements 33
2.4.2 FRA Measurements 34
2.4.3 Discussion on FRA and SCI 38
2.4.4 Summary on FRA and SCI Methods 38
2.5 Conclusion 38
Chapter 3 Transformer Winding Parameters 39
3.1 Introduction 39
3.2 Self and Mutual Inductances of Transformer Winding (Analytical Approach) 40
3.2.1 Self - Inductance 43
3.2.2 Mutual – Inductance 44
3.3 Series and Shunt Capacitances of Transformer Winding 46
3.3.1 Series Capacitance 46
viii
3.3.2 Shunt Capacitance 51
3.4 Verification of Calculated Parameters Using Manufactured Model Transformer 51
3.4.1 Manufactured Model Transformer (Test Object) 51
3.4.2 Inductance Calculation of the Test Object 53
3.5 Conclusion 56
Chapter 4 Transfer Function Model of Air-Core Transformer Winding 57
4.1 Introduction 57
4.2 Modelling 57
4.3 Discussion on Resonant Frequencies 63
4.4 Verification of Mathematical Calculation Using Practical Measurement 64
4.5 Practical Study 66
4.5.1 Case Study 1 66
4.5 Conclusion 69
Chapter 5 Low Frequency Interpretation of FRA Signature 71
5.1 Introduction 71
5.2 Flux Division Theory 72
5.2.1 Technical Concept 72
5.2.2 Flux Division Measurement (FDM) 73
5.3 Mathematical and Practical Approach to Interpret Low-frequency Band 76
5.3.1 General Interpretation of FRA Trace 76
5.3.2 Practical Approach 79
5.4 Mathematical Approach 83
5.5 Effect of Core Configuration on FRA Trace 89
5.6 Shunt Capacitance Influence 91
5.6.1 Practical Approach 92
5.6.2 Physical and Mathematical Approach 93
5.7 Conclusion 95
Chapter 6 Axial and Radial Deformation of Transformer Winding 97
6.1 Introduction 97
6.2 Axial Deformation and Its Impacts on Winding Parameters 98
6.2.1 Mutual Inductance of Circular Filaments Whose Axes Are Inclined to One Another98
6.2.2 Capacitances of Circular Filaments Whose Axes Are Inclined to One Another 103
6.3 Radial Deformation and Its Impacts on Winding Parameters 104
6.3.1 Self and Mutual Inductances in Radial Deformation 104
6.3.2 Series and Shunt Capacitances in Radial Deformation 106
6.4 Numerical Example 107
6.4.1 Axial Deformation of a Disk 107
6.4.2 Radial Deformation along the Winding 111
6.5 A Summary on Axial and Radial Deformations 112
6.6 FRA Simulation Study and Practical Measurement Results 113
6.6.1 Inductance Variation (Simulation) 113
6.6.2 Inductance Variation (Practical Study) 115
6.6.3 Shunt Capacitance Variation (Simulation) 117
ix
6.6.4 Shunt Capacitance Variation (Practical Study) 119
6.6.5 Series Capacitance Variation (Simulation) 121
6.6.6 Series Capacitance Variation (Practical Study) 123
6.6.7 Resistance Variation (Simulation) 125
6.6.8 Resistance Variation (Practical Measurement) 127
6.6.9 Conductance to Ground (G) Variations (Simulation) 130
6.6.10 Turn-to-Turn Conductance (g) Variations (Simulation) 132
6.6.11 Conductance Variation (Practical Study) 133
6.7 Conclusion 133
Chapter 7 Temperature and Moisture Content Influences on FRA Signature 135
7.1 Introduction 135
7.2 Transformer Water Dynamic 135
7.2.1 Transient 136
7.2.2 Equilibrium 137
7.3 Practical Study 138
7.3.1 Test Object and Setup 138
7.3.2 Case Study 1 (‘Wet’ Model Transformer) 139
7.3.3 Discussion 1 144
7.3.4 Case Study 2 (‘Dry’ Model Transformer) 149
7.3.5 Discussion 2 151
7.3.6 Case Study 3 (Three-Phase Transformer) 153
7.4 Verification of Practical Results Using Modelling and Simulation 154
7.5 Influence of Temperature and Moisture Content on FRA Statistical Indicators 160
7.6 Practical Solution to Modify Statistical Indicators 161
7.7 Transformer Winding Dry-out Influence on Frequency Response Trace 165
7.8 Conclusion 167
Chapter 8 On-line Transformer Winding Deformation Diagnosis 169
8.1 Introduction 169
8.2 Advanced Methods in On-line Transformer Winding Deformation Diagnosis 170
8.2.1 Vibration Method 170
8.2.2 Communication Method 170
8.2.3 Current Deformation Coefficient Method 171
8.2.4 Ultrasonic Method 171
8.2.5 Online Short Circuit Impedance and Winding Stray Reactance Method 171
8.2.6 On-line Frequency Response Analysis (On-line FRA) 173
8.3 Discussion 175
8.4 Problem Statement on Online FRA Setup 179
8.5 Challenges with On-line FRA Setup 180
8.6 Case Studies 181
8.6.2 Test Procedure 182
8.6.3 Interpretation 183
x
8.7 Discussion 191
8.8 Conclusion 192
Chapter 9 Conclusion and Future Research 193
9.1 Conclusion 193
9.2 Future Research 196
9.2.1 FRA Test Setup Development 196
9.2.2 Transformer Humidity Recognition Using FRA 196
9.2.3 Transformer Dry-out Assessment Using FRA 197
9.2.4 Oil and Paper Insulation Aging and Oil Replacement 198
9.2.5 On-line Transformer Winding Deformation Recognition 198
Appendix A Developed Software to Calculate Statistical Indicators 199
A.1 Introduction 199
A.2 Implemented Statistical Indicators 200
A.2.1 Correlation Coefficient (CC) 200
A.2.2 Maximum Absolute Difference (DABS) 200
A.2.3 Absolute Sum of Logarithmic Error (ASLE) 200
A.2.4 Min-Max (MM) 200
A.2.5 Standard Deviation (SD) 201
A.2.6 Spectrum Deviation (σ) 201
A.2.7 Cross Correlation Coefficient (CCF) 201
A.2.8 Relative Factor (R) 201
A.3 Developed Software 203
Appendix B Tables and Formulas for Inductance Calculation 204
B.1 Introduction 204
B.2 Tables and Formulas 204
B.3 Self and Mutual Inductances under Buckling 206
B.3.1 Biot-Savart Law and Inductance Calculation 206
B.3.2 Mutual-Inductance under Buckling 209
B.3.3 Self-Inductance under Buckling 215
B.3.4 Numerical Example 217
Appendix C Calculation of Series Capacitance of Intershield Winding 220
C.1 Introduction 220
C.2 Series Capacitance of Intershield Winding 220
C.2.1 Method to Calculate Total Series Capacitance 223
C.2.2 Application of the Proposed Method 225
Appendix D Glassy Model Transformer 227
D.1 Introduction 227
D.2 Test Object Overview 227
D.3 Winding Photos and Winding Schematic 228
Appendix E Dry-out Process of Model Transformer 235
xi
E.1 Introduction 235
E.2 Model Transformer Dry-out 235
E.3 Procedure 236
Appendix F Recognition of Moisture Content in Transformer 239
F.1 Introduction 239
F.2 Water Content Recognition in Oil-impregnated Paper 240
F.2.1 Karl-Fischer Titration 240
F.2.2 Capacitor Method 247
F.2.3 Paper Sample Method 247
F.2.4 Electrical Methods 249
Appendix G Study on Recommended Solution 251
G.1 Introduction 251
G.2 Case Study 1 251
G.3 Case Study 2 254
G.4 Winding Deformation vs. Temperature and Moisture Influences on FRA Spectrum 257
G.4.1 Influence of Internal Short-circuit 257
G.4.2 Influence of Tank Grounding 258
G.5 Conclusion 260
Bibliography 261
xii
List of Figures
Figure 2.1 Winding deformation (Buckling), (a) Free buckling (top view), (b)
Forced buckling (top view), (c) Free buckling (side view) 16
Figure 2.2 Winding deformation, (a) Before tilting, (b) After tilting, (c) Bending
(side view), (d) Bending (close view) 17
Figure 2.3 Transformer transportation schematic 18
Figure 2.4 Transformer transportation, (a) Truck, (b) Rail road, (c) Sea carrier, (d)
Air carrier 21
Figure 2.5 Schematic model of primary, secondary and leakage inductances of a
transformer 22
Figure 2.6 Schematic of transformer core and windings 23
Figure 2.7 Short circuit impedance measurement setup, (a) Single phase
transformer, (b) Three-phase transformer 23
Figure 2.8 Transfer function measurement techniques 24
Figure 2.9 Low, mid and high frequency bands of a typical FRA spectrum
(measured on 400 MVA transformer) 26
Figure 2.10 FRA test setups, (a) End-to-end measurement, (b) Inductive inter-
winding measurement, (c) Capacitive inter-winding measurement, (d)
End-to-end short-circuit measurement 29
Figure 2.11 FRA test setups (detailed connections) 30
Figure 2.12 The measured and fingerprint frequency response magnitudes for
phases A, B and C of the transformer HV side 36
Figure 2.13 Buckled HV winding of phase B, (a) Side view of the middle disks, (b)
Front view of the upper disks 37
Figure 3.1 Air-core transformer winding model 43
Figure 3.2 Coaxial circular conductors, (a) Coaxial filaments, (b) Coaxial disks 45
Figure 3.3 The overall layout of a layer winding including equivalent capacitance
network 46
Figure 3.4 Continuous disk winding schematic 47
Figure 3.5 Equivalent capacitance network of the continuous disk winding 48
Figure 3.6 Pair of disks, cross-section overview and voltage distribution along disks
pair 49
Figure 3.7 The interleaved disk winding 50
Figure 3.8 The intershield disk winding 51
Figure 3.9 Manufactured glassy model transformer (a) Bird’s-eye view, (b) Side
view 52
Figure 3.10 Measured value for the inductance of HV winding 53
Figure 3.11 HV winding schematic (glassy model transformer) 54
Figure 3.12 Measured value for shunt capacitance, HV winding 55
Figure 3.13 Model transformer schematic, top view 55
xiii
Figure 4.1 Equivalent detail circuit of a transformer winding (dx is the infinitesimal
length of winding), (a) Entire winding schematic, (b) Close view of the
conductors and modeled parameters 58
Figure 4.2 Common FRA test setup, (a) For single winding (b) For Transformer 61
Figure 4.3 Simulated and measured frequency response traces for HV winding of
the manufactured glassy transformer 65
Figure 4.4 Continuous and interleaved disk windings, (a) Continuous winding, (b)
Interleaved winding 67
Figure 4.5 Frequency response traces of continuous and interleaved disk windings 67
Figure 4.6 FRA traces of HV windings phases A for 45 MVA (continuous disk
winding) and 66 MVA (interleaved disk winding) power transformers 69
Figure 5.1 Magnetic flux flow due to the transformer winding phase A excitation
when short-circuit occurred in phase B 74
Figure 5.2 Five-limb transformer active part as well as its magnetic circuit 75
Figure 5.3 Magnetic flux division in five-limb transformer due to phase a/A
excitation, short circuit is occurred phase b/B 75
Figure 5.4 Frequency response traces of phases a, b and c 77
Figure 5.5 Equivalent magnetic circuit of three-phase transformer 78
Figure 5.6 FRA measurement setup for phase a when A and C are short-circuited 80
Figure 5.7 Equivalent magnetic circuit for FRA measurement setup of phase a when
the terminals A and C are short-circuited 80
Figure 5.8 Reference and measured FRA traces for phase a when limb C is
surrounded by short circuit loop 80
Figure 5.9 FRA measurement setup for phase a when B and C are short-circuited 81
Figure 5.10 Equivalent magnetic circuit for FRA measurement setup of phase a when
the terminals B and C are short-circuited 82
Figure 5.11 Reference and measured FRA traces for phase a when limb B
surrounded by short circuit loop 82
Figure 5.12 Frequency response traces of phase a (open circuit) and deliberate short
circuit on phase B and C 83
Figure 5.13 Equivalent magnetic circuit of transformer when frequency response
trace is measured for phase a, (a) Normal three-phase, (b) HV winding
phase B is short-circuited, (c) HV winding phase C is short-circuited 86
Figure 5.14 Original and generated frequency response traces of phase a 88
Figure 5.15 Equivalent magnetic circuits when deliberated short circuit is applied for
various limbs 90
Figure 5.16 Frequency response traces for test setups of Figures 5.15 (b) and (h) 91
Figure 5.17 Frequency response traces for test setups of Figures 5.15 (c), (d), (f) and
(k) 91
Figure 5.18 Frequency response traces for phase b when HV side phase B is left open
circuit, short-circuited, short-circuited and grounded 92
Figure 5.19 Active part and related shunt capacitances, (a) Active part upper view
schematic, (b) Shunt capacitance configuration for b where HV side
phase B is just short-circuited and isolated from the ground, (c) Shunt
capacitance configuration for b where HV side phase B is grounded 94
Figure 6.1 Symmetrical axial deformation of a disk, (a) Axial deformation pattern,
(b) Deformed disk 100
Figure 6.2 Asymmetrical axial deformation of a disk, (a), Axial deformation pattern,
(b) Deformed disk 101
xiv
Figure 6.3 Cross-section overview and voltage distribution along deformed winding 103
Figure 6.4 Radial deformation schematic (free-buckling) 104
Figure 6.5 Radial displacement pattern of a disk 105
Figure 6.6 Axially deformed interleaved winding 110
Figure 6.7 Reference and measured frequency response spectra for interleaved
windings 110
Figure 6.8 The modelled winding through finite element (r1= 280 mm, r2 = 400 mm,
r'= ±50 mm, φ= π/2) 112
Figure 6.9 FRA simulation results of winding due to the inductance reduction for 20
and 40 %, (a) Entire FRA spectrum, (b) Expanded view of dash-line
rectangle in Fig. 6.9(a) 114
Figure 6.10 FRA test setup on HV side, (a) Open circuit on LV side, (b) Short circuit
on LV side 116
Figure 6.11 Frequency response of HV winding for open circuit and short-circuited
LV winding 116
Figure 6.12 FRA simulation results of winding due to the shunt capacitance
reduction for 20 and 40 %, (a) Entire FRA spectrum, (b) Expanded view
of dash-line rectangle in Fig.6.12 (a) 118
Figure 6.13 FRA test setup for HV winding, where LV side is short-circuited, and test
object tank is isolated from the ground 119
Figure 6.14 Active part and related shunt capacitances, (a) Shunt capacitance
configuration for HV, where LV side is short-circuited and transformer
tank is grounded, (b) Shunt capacitance configuration for HV, where LV
side is short-circuited and transformer tank is isolated 120
Figure 6.15 Frequency response of HV winding of single phase transformer, LV
winding is short-circuited and also LV winding is short-circuited,
transformer tank is grounded and isolated 120
Figure 6.16 FRA simulation results of winding due to the series capacitance increase,
(a) Entire FRA spectrum, (b) Close-up view of dash-line rectangle in
Fig.6.16 (a), (c) Close-up view including some resonance frequencies 123
Figure 6.17 FRA measurement results of faulty transformer for phase b and c, (a)
Entire FRA spectrum, (b) Close-up view of dash-line circle in Fig.6.17 (a) 124
Figure 6.18 FRA simulation results of winding due to winding resistance increment,
(a) Entire FRA spectrum, (b) Close-up view of region ‘b’ , (c) Close-up
view of region ‘c’ , (d) Close-up view of region ‘d’ 127
Figure 6.19 FRA measurement results of winding due to the resistance increment,
(a) Entire FRA spectrum, (b) Close-up view of region ‘b’ at very low
frequencies, (c) Close-up view of region ‘c’ at high frequency resonant,
(d) Close-up view of region ‘d’ at very high frequencies 130
Figure 6.20 FRA Simulation results of winding due to the conductance to ground (G)
increment, (a) Entire FRA spectrum, (b) Close view of the first resonance
point 131
Figure 6.21 FRA Simulation results of winding due to the turn-to-turn conductance
(g) increment, (a) Entire FRA spectrum, (b) Close view of the first
resonance point 133
Figure 7.1 Water dynamic in paper and oil insulation for different temperatures T1
and T2 (T1 < T2), WCO and WCP; t denotes the time 136
Figure 7.2 Manufactured glassy air-core transformer (setup preparation to study
temperature and moisture impact) 138
Figure 7.3 FRA test setup to examine temperature and moisture variation 139
xv
Figure 7.4 FRA spectra for ‘wet’ model transformer, HV side (a) Entire trace for 30
and 90 °C, (b) Close-up view of region 1 shown by dash-line in Fig. 7.4(a),
frequency band 800 kHz-3 MHz, (c) Close-up view of region 2 shown by
dash-line in Fig. 7.4(a), frequency band 3 MHz – 10 MHz 142
Figure 7.5 FRA spectra for ‘wet’ model transformer, LV side (a) Entire trace for 30
and 90 °C, (b) Close-up view of region 1 shown by dash-line in Fig. 7.5(a),
frequency band 500 kHz - 3.5 MHz, (c) Close-up view of region 2 shown
by dash-line in Fig. 7.5(a), frequency band 3.5 MHz – 10 MHz 144
Figure 7.6 Deviation of total capacitance for HV and LV windings (average deviation
at 90 °C is 6.22 %) 147
Figure 7.7 Moisture content of oil and paper (wet model transformer) 148
Figure 7.8 Vacuum process of the model transformer 149
Figure 7.9 FRA spectra for ‘dry’ model transformer, HV side 150
Figure 7.10 FRA spectra for ‘dry’ model transformer, LV side 150
Figure 7.11 Moisture content of oil and paper (dry model transformer) 151
Figure 7.12 HV winding spectra at 10 and 60 °C (1.6 MVA transformer) 154
Figure 7.13 Schematic of the paper content and the spacer coverage in the insulation
duct 158
Figure 7.14 Simulated FRA spectra for model transformer (shunt capacitance
deviation) 160
Figure 7.15 The chart on preliminary calculation on FRA traces to distinguish
insulation deviation from winding deformation; Xi and Yi were defined in
Chapter 2 164
Figure 7.16 Glassy model transformer 165
Figure 7.17 FRA traces for glassy test object (a) HV winding spectra before and after
dry-out (frequency band 3 kHz-20 MHz, 0 dB < 3 kHz), (b) LV winding
spectra before and after dry-out (frequency band 6 kHz- 20 MHz, 0 dB <
6 kHz). The measurements have been performed for oil-filled model
transformer 166
Figure 8.1 Two-port network 172
Figure 8.2 Side cut-off of a capacitive bushing 174
Figure 8.3 Paralleled impedance with bushing tap (test tap) on phase U 175
Figure 8.4 Wye circuits for on-line frequency response measurement 176
Figure 8.5 Delta circuits for on-line frequency response measurement 176
Figure 8.6 On-line FRA setup for a transformer interleaved winding 181
Figure 8.7 Close view of test setup 182
Figure 8.8 Frequency response traces of normal and defected winding, off-line
setup 183
Figure 8.9 Frequency response traces of normal and defected winding, on-line
setup 183
Figure 8.10 FRA test setup behaviour with and without bushing connection, (a)
Common FRA test setup, (b) FRA test setup with bushing connected, (c)
FRA test setup with bushing connected (low-frequency behaviour) (d)
FRA test setup with bushing connected (mid-frequency behaviour), (e)
FRA test setup with bushing connected (high frequency behaviour) 185
Figure 8.11 Frequency response traces of normal and defected winding, off line and
on-line setup 186
Figure 8.12 Frequency response traces of interleaved (off-line setup and on-line
setup through one coupling capacitor) 187
xvi
Figure 8.13 Paralleled coupling capacitors on conventional bushing 188
Figure 8.14 Frequency response traces of interleaved (off-line setup and on-line
setup through coupling capacitor) 189
Figure 8.15 Frequency response traces of continuous disk winding (off-line setup
and on-line setup through coupling capacitor) 189
Figure 8.16 Frequency response traces of single phase transformer (off-line setup
and on-line setup through coupling capacitor) 190
Figure A.1 Developed software to calculate statistical indicators, snapshot 203
Figure B.1 Biot-Savart law 206
Figure B.2 Magnetic flux determination for circular filament 207
Figure B.3 Concentric circular filaments 208
Figure B.4 Concentric circular filaments, inward buckling demonstration for the
second loop 210
Figure B.5 Concentric circular filaments, demonstration of inward buckling for both
loops 212
Figure B.6 Magnetic flux determination for non-circular filament 215
Figure C.1 Equivalent RLC network for intershield disk winding 221
Figure C.2 Manufactured windings from top to bottom: Interleaved, Continuous,
Intershield with one shield turn, and Intershield with six shield turns in
each disk 222
Figure C.3 Overall scheme and the sample point of voltage for the simulated
winding, (S.P=Sample Point) 224
Figure D.1 Manufactured glassy test object, (a) Side view of winding, (b) Side view
without oil, (c) Entire view, (d) Side view with oil 228
Figure D.2 HV technical drawing, (a) HV technical schematic, (b) Specifications, (c)
Backward step drawing 231
Figure D.3 LV technical drawing, (a) LV technical schematic, (b) Specifications, (c)
Backward step drawing 234
Figure E.1 Dry-out equipment, Dry-out equipment, (a) Wet silica gel, (b) Half dried
silica gel, (c) Silica gel dry-out process (oven view), (d) Dried silica gel
(oven view), (e) Dried silica gel, (f) Silica gel container to break the
vacuum after oil dry-out process, (g) Oil circulator, (h) Internal piping of
oil circulator, (i) Oil circulator vacuum gauge, (j) Transformer oil, (k)
Vacuum pump, (l) Internal motor pump of circulator, (m) Glassy anti-
vacuum container of oil circulator, (n) Karl-Fischer equipment, (o) Metal
stopcock, (p) Vacuum gauge for test object dry-out, (q) Model
transformer dry-out process (oven view) 238
Figure F.1 Oil sample containers, (a) Glass laboratory bottle, (b) Glass syringe 241
Figure F.2 Oil sample containers, (a) Side view, (b) Front view 242
Figure F.3 KFT equipment, (a) Digital micro scale, (b) Glassy container, main unit
and keyboard 243
Figure F.4 KFT unit main compartments, (a) Double platinum wire electrode (0.8 x
4 mm), (b)Generator electrode for Karl Fischer titrations, with
diaphragm, (c) KF absorber tube for coulometer cell, (d) SGJ stopper, (e)
Stopper, (f) Plastic tube, (g) Titration vessel holder for coulometric cells,
(h) KF titration vessel, (i) Keyboard for 756 KF Coulometer 244
Figure F.5 Oommen equilibrium curves for oil-paper system 246
Figure F.6 MIT-developed curves for water equilibrium in cellulose/mineral oil
systems 246
xvii
Figure F.7 Capacitance sensor layers 247
Figure F.8 Dean-Stark apparatus, (a) Condenser, (b) Receiver, (c) Flask, (d)
Assembled parts 248
Figure F.9 Different views of measuring water content of the sampled papers using
Dean-Stark method 248
Figure F.10 Variation of loss factor of oil-immersed paper with frequency and
dominant influences 250
Figure G.1 Reference and derivative spectra for HV winding of glassy model
transformer, (a) at 30 °C, (b) 90 °C 252
Figure G.2 Frequency response spectra for HV winding when the LV winding was
left open and short-circuited (test object without oil), (a) Entire
frequency band (20 Hz – 20 MHz), (b) The area enclosed by dash-line
rectangle in Fig. G.2(a), (100 kHz – 20 MHz) 255
Figure G.3 HV winding spectra (a) Reference spectrum and its derivative (100 kHz
– 20 MHz), (b) Measured spectrum and its derivative (100 kHz – 20
MHz) 256
Figure G.4 (a) HV winding frequency response spectrum when LV winding is open-
circuited (original spectrum) and short-circuited (affected frequency-
band 20 Hz-3 MHz), (b) FRA spectra of HV winding for isolated and
grounded tank (affected frequency-band 300 kHz-20 MHz), (c) FRA
spectra for HV winding due to moisture migration from paper into the oil
insulation at 30°C and 90°C, re-plotted from Fig. 7.4 for comparison
(affected frequency-band 800 kHz-20 MHz) 259
xviii
List of Tables
Table 2.1 FRA measurement connections (phases A, B and C) 30

Table 2.2 CC and SD criteria 32
Table 2.3 Level of deformation based on Rxy 32
Table 2.4 400 MVA Step-up transformer parameters 34
Table 2.5 Measured SCI Values 34
Table 2.6 CC, SD values 36
Table 2.7 Rxy values 36
Table 4.1 Transformer specifications 68
Table 5.1 Transformer specifications 78
Table 5.2 FRA setup connections for the configuration in Fig. 5.15 90
Table 6.1 Calculated capacitance between the winding and metal container (tank) 112
Table 6.2 Technical specifications of single phase test object 115
Table 7.1 Ambient air relative humidity and moisture in paper 140
Table 7.2 HV winding electrical parameters for 30 and 90 °C 146
Table 7.3 LV winding electrical parameters for 30 and 90 °C 146
Table 7.4 HV winding capacitance ratio, anti-resonance and resonance frequencies
for quoted points in Fig.7.4(b) and Fig. 7.4(c ) 147
Table 7.5 LV winding capacitance ratio, anti-resonance and resonance frequencies
for quoted points in Fig.7.5(b) and Fig. 7.5(c) 148
Table 7.6 HV winding capacitance ratio, anti-resonance and resonance frequencies 151
Table 7.7 LV winding capacitance ratio, anti-resonance and resonance frequencies 151
Table 7.8 FRA deviation and total capacitance variation for 0.5 % WCP change 152
Table 7.9 Measured DDF at 5 kV before and after dry-out process 153
Table 7.10 HV winding capacitance ratio, anti-resonance and resonance frequencies
for quoted points in Fig. 7.12 154
Table 7.11 Calculated statistical indices 160
Table 7.12 Capacitance ratio, anti-resonance and resonance frequencies for quoted
points in Fig. 7.16(a) 167
Table 7.13 Capacitance ratio, anti-resonance and resonance frequencies for quoted
points in Fig. 7.16(b) 167
Table 8.1 Typical values for bushing capacitances 174
Table B.1 Values of KN for single layer coil 204
Table B.2 Values of R0 for inclined circles 205
Table B.3 Values of F for parallel circles, rd/2R=Λ 205
Table C.1 Initial voltage distribution (%) 225
Table G.1 Calculated values for Z{î }n and Z{î }m 253
Table G.2 Calculated values for reference spectrum Z{î }n, and measured spectrum
Z{î }m 257
xix
Chapter 1. Introduction
Chapter 1
Introduction
1.1 Problem Statement
Power transformers are put in service under different environmental, electrical and
mechanical conditions, and may be subject to enormous hazards during the course of
operation. They are commonly considered as the heart of the transmission and distribution
sectors of electric power systems. Thus, monitoring their condition and diagnosing faults
are important parts of the maintenance function. Utility engineers strive to keep power
transformers in service, and to prevent even short-term outages. Failure of a transformer
can cause extensive damage to connected equipment owned by consumers and/or the
utility.
Mechanical defects are probably the most common cause of problems in transformers
which can put this asset out of service for a long time. They may be due to short circuit
currents, earthquakes, careless transportation between sites, explosion of combustible
gases accumulating in the transformer oil, etc. In fact, transformer internal short circuits
will adversely affect transformer insulation system which includes paper and pressboard,
while external short circuits in power networks will subject the transformer winding to
substantial electro-dynamic forces. Direction of forces is perpendicular to the magnetic
field vector according to electromagnetic theorem. They cause winding deformation in the
axial and/or radial directions, hoop buckling, tilting, spiraling, displacements between high
and low voltage windings, shorted or open-circuited turns, partial winding collapse,
loosened clamping structures, core movement, faulty grounding of core or screens, broken
clamping structures, and intermittent internal connections. Repair of such defects may
require taking the transformer out of service, which could prove costly to the utility. To
recognize transformer mechanical defect without opening the transformer tank and also
avoid undesirable costly maintenance, the method of Frequency Response Analysis (FRA)
has been introduced and widely employed in the power industry. FRA is considered a
highly accurate, fast, economical and non-destructive method of detecting winding defects
and damage in the transformer core. This method has been studied for many years, while
1
FRA data (trace) interpretation is still under development. On the other hand, there has
been a concern that the changes in transformer insulation characteristics, in particular the
temperature and moisture content, can influence FRA results.
Hence, not only interpretation of FRA data due to the winding deformation has still
remained as a challenge, but also FRA trace alteration due to the temperature and moisture
content variation should be clarified.
Furthermore, another challenge that is worthwhile to tackle is to be able to distinguish
whether the measured FRA deviations are indicative of transformer winding deformations
or due to insulation characteristic changes.
1.2 Motivation
The IEC Standard 60076-18, Ed.1 [1], on FRA test technique points out that not only FRA
traces is affected by winding deformation, but also it could be affected by the temperature
changes. A figure showing deviated FRA trace from the reference value is provided in this
standard. However, deviation amounts in resonance peaks as well as their causes and
effects are not discussed.
In addition, IEEE Standard C57.149 [2] on application of frequency response analysis for
oil-immersed transformers mentions that “large temperature difference, typical much more
than 10 °C, between two measurements will slightly influence the response at higher
frequencies”, though, further discussion is not provided in this standard either.
The research undertaken in this thesis targets the IEC and IEEE Standard comments by
developing more comprehensive and flexible transformer models that will facilitate a
profound study on this issue.
2
1.3 Current State of the Art
The AC impedance or admittance of any RLC network is a function of frequency. Dick and
Erven have utilized this idea [3] in transformer diagnosis and introduced Frequency
Response Analysis (FRA). Nowadays FRA measurement is widely accepted as an effective
diagnosis approach in transformer mechanical integrity investigation [4]-[7].
The initial FRA measurement during transformer factory tests is known as the transformer
winding fingerprint (baseline or signature) [8]-[9]. The distributed resistance, capacitance
and inductance of the transformer winding in the frequency range 20 Hz–2 MHz determine
the shape of the reference frequency response signature. These parameters are influenced
by not only the material electrical characteristics but also the structural geometry. Thus,
transformer geometric structure changes will lead to frequency response trace deviation
and winding deformation recognition, accordingly [10].
Since the last decade, there has been an interest in progressing the diagnostic benefits
associated with FRA. Also different approaches have been tried to improve FRA
interpretation ability. Based on literatures, studies in this area can be categorized into five
major groups: FRA classification, FRA development, FRA assessment, FRA interpretation
and on-line FRA application.
In the case of FRA classification, the literatures are trying to explore transformer winding
deformation type through measured FRA spectrum [11]-[14]. For instance, Rahimpour et al
[12], have determined and classified mathematical coefficients to explore winding axial
deformation through deviated FRA trace.
In the case of FRA development, researchers have tried to extend the diagnostic potential of
the FRA method [15]-[19]. They believe not only FRA can be utilized as a method to explore
transformer mechanical integrity but also it can provide significant information about
transformer internal condition. Hence, FRA might eventually replace some conventional
methods such as turn ratio, flux division, vector group and short circuit impedance
measurements.
Regarding FRA assessment, the literatures have concentrated on transformer winding
deformation diagnosis through measured FRA spectrum [20]-[26]. In fact, the main
concern of these studies has focused on using statistical indicators to recognize winding
deformation or active part displacement apart from deformation type. Researchers believe
that such an approach can lead to development of an intelligent system for FRA
investigation and diagnosis of transformer mechanical condition.
The next category, namely FRA interpretation, is concerned with the interpretation of FRA
fingerprint [27]-[29]. The correlation between FRA spectrum oscillations and transformer
3
physical configuration has been discussed. It has been emphasized that physical
interpretation of FRA spectrum will lead to a methodical and hence better understanding of
transformer physical condition.
The last category, namely on-line FRA application, is focused on real-time FRA
measurement under normal operating conditions. The literatures in this area [8], [10],
[30]-[37] believe that since on-line high voltage monitoring systems are under
development now, all of off-line diagnosis methods need to be replaced with on-line
application. Therefore, the potential problems of adapting off-line FRA to on-line situations
need to be identified and addressed.
Apart from discussed categories, a number of researchers have made significant
contributions in transformer modeling to support FRA interpretation [38]-[43]. These
literatures contain valuable mathematical calculations. Transformer active part modeling
through different approaches has been elaborated to study frequency response. Other
studies in this field show interest on FRA circuit setup, connections as well as
environmental impacts on frequency response trace deviation [44]-[50].
Despite all attempts there are many aspects in the FRA concept that need to be addressed
and overcome. One of those aspects is the interpretation of resonances and anti-resonances
within FRA trace for each and every transformer. Previous efforts should be further
developed with supporting theory to help interpret FRA fluctuations.
In addition, as discussed earlier, changes in winding configuration would almost certainly
cause changes in the frequency response trace. On the other hand, transformer oil
deterioration as well as water absorbed by the paper insulation cause transformer
insulation characteristics to change over the time [51]-[52]. In fact, insulation characteristic
can change due to a number of factors. Temperature, humidity, oil acidity, oil interfacial
tension, oil contamination, oil viscosity, oil breakdown voltage and degree of
polymerization in paper are major factors to be considered, to name a few. Any change in
one or more of these factors will be reflected in some change in winding electrical and
magnetic behaviour. Among the factors, water content and temperature variation would be
more significant.
In fact, capacitances, self and mutual inductances and conductor resistances might be
altered due to any changes in above mentioned factors. In turn, the frequency response of
the winding will change accordingly. The concern on the influence of insulation
characteristic on the frequency response trace becomes significant when this trace is
required to be interpreted for diagnostic purposes. FRA deviation due to the winding
deformation or insulation characteristic changes turns into a more substantial challenge
4
when FRA baseline and measured traces are taken at different temperatures and moisture
contents of the transformer winding. In this regard, investigation of the effect of changes in
complex dielectric permittivity of transformer insulation on its high frequency response is
the primary aim of a study by Abeywickrama et al [16]. In this study, a single-phase
transformer without a core was first modeled and then developed to three-phase by means
of a lumped parameter circuit to illustrate the insulation impacts on frequency domain
response [16]. This study concludes that FRA technique could be implemented for
diagnostics of the insulation quality in power transformers but more work is necessary for
establishing reliable interpretation.
The impact of oil and temperature on the initial voltage distribution for air and oil
immersed transformer layered windings has been studied by Florkowski et al [53]. This
investigation was subsequently extended for different disk windings in [54] by the same
authors. The oil and temperatures influences on the frequency characteristic have been
experimentally investigated and discussed in both works.
Reykherdt et al [19] studied the sensitivity of the FRA signatures to winding temperature
and moisture content. However, the reason for changes in the FRA spectrum was not
rigorously discussed in detail.
The influence of moisture ingress on transformer winding distributed parameters was
mentioned in a study by Abu-Siada et al [55], while the FRA method as a technique for
moisture diagnostics of power transformers was elaborated by Yadav et al [56]. However,
the abstract of the latter was too promising in relation to what was actually reported in the
article.
A study by Bjerkan [57] claims that the transformer insulation aging will impact the FRA
signature, but this impact is not significant. The IEC Standard 60076-18, Ed.1 [1] provides
information of deviated FRA spectra due to just temperature changes. This standard also
clearly states that “temperature affects the frequency response” but supporting discussion
is not provided.
Hence, a comprehensive study on FRA trace alteration due to the temperature and
moisture content changes is necessary to accurately quantify the FRA signatures variation
under such circumstances. Statistical indicators as available FRA evaluation methods
should be also examined once temperature and moisture content are changed. Correlation
between the moisture content in paper insulation and FRA trace variation can also be
considered as another important research area.
5
1.4 Research Objectives
Based on aforementioned discussion, it appears that a number of key areas are in need of
further research.
At first, the main reasons of resonances and anti-resonances of FRA signature should be
clarified through the mathematical modeling of transformer winding. Next, it would be a
novel idea to fabricate a model transformer as an experimental test object to study
mechanical integrity as well as temperature and moisture variation on its FRA signature,
synchronously. This manufactured model transformer can also be utilized for mathematical
model verification. The verified model can then be used for studying the axial and radial
deformation of transformer windings of FRA signature. In addition, it could be utilized to
investigate the insulation characteristics influences on FRA data. Based on this, the
objectives of this thesis are:
 Transformer winding modeling to interpret the resonances and anti-resonances in
FRA trace,
 Manufacturing a model transformer as a unique test object to experimentally verify
the developed model and studying the mechanical integrity,
 Study on axial and radial deformation of transformer winding as the main roots of
any transformer winding deformation category,
 Explore the influence of temperature and moisture contents on FRA signature and
compare the results to those coming through mechanical defects,
 Estimation of moisture variation through FRA trace,
 Development of a comprehensive software including all available FRA evaluation
indicators to investigate the impact of FRA trace deviation due to the temperature
and moisture variation on indicators,
 Investigate and suggest possible solution to distinguish the winding deformation
from insulation characteristic impacts on FRA spectrum.
6
1.5 Thesis Overview
This study is focused on the influence of different origins on the transformer winding FRA
signature. Accurate recognition and interpretation of these origins are crucial for
developing a reliable transformer condition monitoring system. The different reasons
causing FRA trace deviations are going to be examined theoretically and through practical
experiments. To satisfy this objective, the thesis is structured in the following manner.
Chapter 2 provides a review on transformer winding deformation. It also discusses
available off-line methods for transformer mechanical defects recognition. The
fundamentals of frequency response analysis as the main subject of this study are
introduced and through a practical approach compared to the conventional method of
Short Circuit Impedance (SCI) measurement. FRA evaluation methods based on statistical
indicators are then discussed in detail to support the work in later Chapters.
Chapter 3 derives the self and mutual-inductance of an air-core transformer winding in
detail. To do this, literatures including some over a hundred years old have been taken into
account. In addition, calculations of the series and shunt capacitance of the transformer
winding are discussed. At the end, to verify the presented analytical approach on winding
parameters, a numerical example is provided and the calculated parameters based on the
analytical approach are compared to the measured parameters in manufactured test object.
Chapter 4 develops a mathematical model for air-core transformer winding to account
for oscillations in the FRA trace. This model is verified through FRA measurement on a
model transformer manufactured. This in turn enables interpretation of the mid-frequency
oscillations once the transformer core is not taken into consideration.
Chapter 5 examines the FRA low-frequency band. Through practical experiments, the
main resonances and anti-resonances in FRA trace for star connection of the transformer
windings are analyzed and discussed. The technique developed in this Chapter can be
utilized for transformer core defects recognition.
Chapter 6 discusses specifically on axial and radial deformations in transformer
winding. The mathematical foundation developed in this Chapter is believed to be essential
for future researches on winding deformation. A small transformer winding is considered
as a simple model and winding parameters are derived in detail for each and every
deformation category. The results are then used for simulation of different winding
deformation category. In the case of radial deformation, through a simple winding
deformation, the results achieved by the analytical approach are compared to Finite
Element Method (FEM) results.
7
Chapter 7 experimentally discusses the effect of temperature and moisture content

variation on FRA trace. Accurate measurements are performed on the manufactured test
object and the influence of moisture migration on FRA data is derived. Practical results are
then validated in this Chapter through the modeling and simulation and the main reason of
FRA deviation due to insulation characteristics changes are derived. FRA evaluation
methods are examined under such circumstances. This Chapter also opens up discussion on
recognition of moisture content variation in paper insulation using FRA. In addition, a
solution to distinguish the transformer winding deformation from insulation characteristic
impacts on FRA spectrum is also discussed in this Chapter.
Chapter 8 introduces the online FRA measurement as a possible solution to distinguish
winding deformation from insulation characteristic effects on the FRA trace. This Chapter
also reviews all recommended online transformer winding deformation recognition
methods and practically compares them with online FRA application. As the last part of the
thesis, this Chapter opens the doors for future research in this regard.
The conclusion to this research is given in Chapter 9 including a summary of the results
achieved and a discussion on future research.
The thesis also includes a number of appendices. Appendix A provides detailed
information of the software developed to calculate the various statistical indicators to for
evaluating FRA. This software is able to analyze the FRA spectra different frequency sub-
bands.
Appendix B gives the tables and formulas for inductance calculation used in Chapters 3
and 6.
Appendix C provides a detailed calculation on series capacitance of intershield winding.
It examines how the difference in the number of shields can produce a significantly
different series capacitance.
Appendix D provides detailed information on the glassy model transformer including
technical drawing. This transformer was specifically designed and fabricated for this
research work.
Appendix E describes transformer dry-out process and specifically discusses the process
used in this thesis for the work required in Chapter 7.
Appendix F gives an overview on moisture content measuring techniques in transformer
and specifically focuses on the Karl-Fischer Titration (KFT) used in Chapter 7.
Appendix G provides a practical example on the solution recommended in this thesis to
distinguish the insulation characteristic changes from the mechanical deformation in their
influence on the FRA trace.
8
1.6 Key Contributions
 Demonstrated analytically how axial and radial deformation in transformer

winding will influence winding parameters,
 Developed a comprehensive model of transformer winding to interpret mid-
frequency resonances and anti-resonances in FRA trace,
 Demonstrated experimentally that low-frequency oscillations in FRA trace are
influenced by the middle and lateral transformer core limbs,
 Demonstrated experimentally that FRA trace can be influenced through
temperature and moisture content variation,
 Examined FRA evaluation indicators when moisture content of transformer paper
insulation is changed,
 Developed estimates for the moisture content variation in transformer paper
insulation through FRA trace study. This work may impact significantly the future
of moisture content recognition in transformer,
 Demonstrated that the effectiveness of the transformer drying process could be
verified by examining the FRA trace. This can serve as a fast, reliable and non-
destructive technique for transformer dry-out evaluation,
 Developed a comprehensive software package to calculate all evaluation indicators
of FRA trace,
 Recommended a technique to distinguish insulation characteristic variation from
winding deformation in FRA trace, also recommended on-line FRA measurement as
another possible solution to distinguish the impact of these phenomena on the FRA
spectrum,
 Discussed the practical challenges in on-line FRA measurement and a technique to
extract maximum information through the online FRA setup recommended.
9
1.7 Publications
Journal Papers
1. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, “Advanced

Transformer Winding Deformation Diagnosis: Moving from Off-line to On-
line”, IEEE Transaction on Dielectric and Electrical Insulation, Vol. 19, Issue. 6,
pp. 1860-1870, 2012.
2. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “Transformer Frequency

Response Analysis: Mathematical and Practical Approach to Interpret Mid-
frequency Oscillations”, IEEE Transaction on Dielectric and Electrical
Insulation, Vol. 20, Issue 6, pp. 1962-1970, 2013.
3. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, Toan Phung

“Frequency Response Analysis and Short Circuit Impedance Measurement in
Detection of Winding Deformation in Power Transformer”, IEEE Electrical
Insulation Magazine, May/June, Vol. 29, no. 3, pp. 33-40, 2013.
4. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “Influence of Temperature and

Moisture Content on Frequency Response Analysis of Transformer Winding”,
IEEE Transaction on Dielectric and Electrical Insulation, Vol. 21, Issue 3, pp.
1393-1404, 2014.
5. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “Axial and Radial Transformer
Winding Deformation: An Analytical Step towards Frequency Response
Understanding”, IEEE Transaction on Dielectric and Electrical Insulation,
(submitted) 2014.
Conference Papers
6. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung, “Practical

Challenges in Online Transformer Winding Deformation Diagnostics”, IEEE
International Conference on Electric Power and Energy Conversion
Systems (EPECS’11), Sharjah, UAE, Nov. 15-17, pp. 1-6, December 2011.
7. Mehdi Bagheri, Mohammad S. Naderi, “Moisture Diagnostics of Power

Transformers Using Dielectric Response Methods and Paper Samples
Method”, IEEE Electrical Insulation Conference (EIC’11), Maryland, U.S.A., pp.
36-40, June 2011.
8. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung, “LED vs.
CFL Effects on Distribution Transformer Efficiency”, 4th International
Engineering Conference (EnCon’11), Kuching, Sarawak, Malaysia, pp. 1-5,
December 2011.
9. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung, “Online

Transformer Winding Deformation Diagnosis: A Profound Insight to
Methods”, 26th International Power System Conference (PSC’2011), Tehran,
Iran, pp. 1-9, Oct and Nov 2011.
10. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung “Bushing
Characteristic Impacts on On-line Frequency Response Analysis of
10
Transformer Winding”, IEEE International Conference on Power and Energy

(PECON’12), Kota Kinabalu, Malaysia, pp. 956-961, December 2012.
11. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, Damin Zhang,

“Transformer Frequency Response Analysis: A Mathematical Approach to
Interpret Mid-frequency Oscillations”, IEEE International Conference on Power
and Energy (PECON’12), Kota Kinabalu, Malaysia, pp. 962–966, December
2012.
12. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung “Dean-
Stark vs FDS and KFT Methods in Moisture Content Recognision of
Transformers”, IEEE International Conference on Power and Energy
(PECON’12), Kota Kinabalu, Malaysia, pp. 712-717, December 2012.
13. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung,

“Frequency Response Analysis to Recognize Inductance Variation in
Transformer Due to Internal Short Circuit”, IEEE International Power and
Energy Conference (IPEC’12), Ho Chi Minh, Vietnam, pp. 1-5, December 2012.
14. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung, “Case
Study on FRA Capability in Detection of Mechanical Defects within a 400MVA
Transformer,” CIGRE, 21, rue d’Artois, F-75008 Paris, France, pp.1-9, August
2012.
15. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung , “FRA vs.
Short Circuit Impedance Measurement in Detection of Mechanical Defects
within Large Power Transformer”, IEEE International Symposium on Electrical
Insulation (ISEI’12), San Juan, Puerto Rico, June 10-13, pp. 301-305, June 2012.
16. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung

“Frequency Response Analysis vs. Flux Division Measurement in Detection of
Transformer Winding Internal Short Circuit”, IEEE International Conference on
Power System Technology (POWERCON’12), Auckland, New Zealand, pp. 1-5,
November 2012.
17. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung

“Transformer Efficiency and De-rating Evaluation with Non-Sinusoidal Loads”,
IEEE International Conference on Power System Technology (POWERCON’12),
Auckland, New Zealand, pp. 1-6, November 2012.
18. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “On-line Transformer

Frequency Response Analysis: Moisture and Temperature Influences on
Statistical Indicators”, IEEE International Conference on Smart
Instrumentation, Measurement and Applications (ICSIMA’13), Kuala Lumpur,
Malaysia, pp. 119-124, November 2013 (Best Paper Award).
19. Mehdi Bagheri, Toan Phung, Trevor Blackburn “Transformer Core Influence
on Frequency Response Spectrum Oscillations”, International Symposium on
High Voltage (ISH’13), Seoul, South Korea, pp. 1772-1777, August 2013.
20. Mehdi Bagheri, B. T. Phung, Trevor Blackburn “Impact of Transformer

Winding Dry-Out on Frequency Response Analysis”, IEEE International
Conference on Electrical Insulation and Dielectric Phenomena (CEIDP’13),
Shenzhen, China, pp. 1113-1116, October 2013.
11
21. Mehdi Bagheri, B. T. Phung, Trevor Blackburn, “Impacts of Transformer

Winding Series Capacitance on FRA Spectrum Oscillations”, CIGR ’13, Int’l
Study Committee Meeting and Colloquium, Brisbane, Australia, pp. 1-9,
September 2013.
22. Mehdi Bagheri, Toan Phung, Trevor Blackburn, Ali Naderian “Shunt
Capacitance Influences on Transformer FRA Spectrum”, IEEE International
Electrical Insulation Conference (EIC’13), Ottawa, Canada, pp. 225-229, June
2013.
23. Mehdi Bagheri, Toan Phung, Trevor Blackburn, Ali Naderian “Influence of
Temperature on Frequency Response Analysis of Transformer Winding”, IEEE
International Electrical Insulation Conference (EIC’13), Ottawa, Canada, pp. 40-
44, June 2013.
24. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “A Practical Solution to

Distinguish Winding Deformation from Insulation Characteristic Impacts on
FRA Spectrum”, International Conference on Condition Monitoring and
Diagnosis (CMD’14), Jeju, South Korea, (submitted) 2014.
25. Mehdi Bagheri, B. T. Phung, Trevor Blackburn, “The Influence of Dielectric

Dissipation Factors on Transformer Frequency Response Analysis”, IEEE
International Conference on Electrical Insulation and Dielectric Phenomena
(CEIDP’14), Des Moines, IA, U.S.A., (submitted) 2014.
26. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “Influence of Moisture Content
Variation on Frequency Response Analysis of Transformer Winding”, IEEE
International Electrical Insulation Conference (EIC’14), Philadelphia, U.S.A.,
(approved for publication) June 2014.
27. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “Paper Moisture Variation vs.
Mechanical Deformation Impacts on Transformer Frequency Response
Spectrum”, 7th International Symposium on Electrical Insulating Materials
(ISEIM’14), Toki Messe, Niigata City, Japan, (approved for publication) June
2014.
12
Chapter 2. Transformer Winding Deformation and Diagnosis Techniques
Chapter 2
Transformer Winding Deformation

and Diagnosis Techniques
2.1 Introduction
This Chapter discusses transformer winding deformation and available diagnosis

techniques. The discussion begins with investigation on the main causes that put
transformers out of service. It then proceeds to highlight the impacts of short circuit on
transformer winding. It specifically debates on radial and axial deformations and then
introduces existing diagnosis methods. Frequency Response Analysis (FRA) as the main
tool of this study and its evaluation methods are elaborated and discussed in detail. The
advantages FRA as compared to other available methods for winding deformation
diagnosis are demonstrated through carefully-designed practical work.
This Chapter is structured in the following manner. Winding deformation is discussed in
Section 2.2 and its diagnosis techniques are described in Section 2.3. Comparison between
FRA and short circuit impedance (SCI) measurement is presented in Section 2.4 with
concluding remarks in Section 2.5.
2.2 Transformer Winding Deformation and Displacement
Different types of transformer windings are designed and manufactured based on voltage
level and electromagnetic relations. Spacers, barriers, dense woods and other materials
are used to provide mechanical support for the winding. Spacers are employed to separate
one disk from another disk in order to provide easy heat dissipation as well as mechanical
support. They can also help as a part of the cooling system by providing appropriate
distance between two adjacent disks for oil flow. Larger space between the disks might
result in oil flow speed reduction and inefficient cooling whereas small space can impede
oil flow and make cooling function useless. Hence, there is an optimum disk space based
on transformer winding design and its physical configuration and also oil viscosity.
13
On one hand, the spacers can be placed in layers to provide suitable disk-to-disk distance;
on the other hand they can be arranged horizontally in a disk to provide mechanical
support. In this regard, the number of spacers per disk (horizontally) is given by:
Nspacer   Dave , (2.1)

s Wspacer
where, Dave is the average diameter of winding, Wspacer is the spacer width, and θs
represents the desirable angle between spacers based on mechanical support concepts.
Transformer mechanical defects can occur due to the many disturbances like as short
circuit currents, severe explosion of combustible gas in transformer oil, earth quake, or
even improper transportation. Winding deformations are reportedly happening between
spacers and barriers. Statistical evidence has shown that well placed spacers can prevent
winding deformation during short circuit, especially for inner winding subjected to the
radial inward forces. In fact, winding sections between spacers are considered the weak
points of a winding during short circuit events.
2.2.1 Short Circuit Current
Of all the possible causes of transformer failure, mechanical deformation of windings as a

result of high short-circuit currents is probably the most common. Such currents may
generate radial, axial or combined forces acting on a transformer winding. The result
could be radial, axial or angular deformation of the windings, or conductor rupture.
Transformer winding deformation may be categorized as follows [18]:
 Radial forces
 Forced buckling
 Free buckling (hoop buckling)
 Hoop tension (stretching)
 Relaxation buckling
 Axial forces
 Tilting (cable-wise tilting, strand-wise tilting)
 Conductor bending between radial spacers
 Combined forces
 Spiralling
 Telescoping
 Twisting
14
In all cases the force F acting on the transformer winding is given by:
F   I sh .d   B (2.2)

where, Ish is the short-circuit current, B is the vector of magnetic induction, and λ is the
winding length.
2.2.1.1 Radial Forces
Radial forces produced by the axial leakage field act outwards on the outer winding
tending to stretch the winding conductors and it can cause hoop stress. On the other hand,
the radial forces cause the inner winding to experience radial compressive stress. The
radial force due to the axial leakage flux in the gap between the two windings is calculated
as follow [10], [58]-[60]:
 20 NI 
Fradial   Dave  ( 2 NI ), (2.3)
 2H w 
 
where, Hw is the winding height, Dave is the average winding diameter, NI denotes R.M.S.
winding’s ampere-turns value, and µo represents vacuum permeability.
Various types of deformations due to the radial forces were discussed earlier. Amongst all
of them, the buckling type of deformation is the most often reported in transformer
windings [10]. Radial forces in concentric windings lead to free buckling and forced
buckling as illustrated in Fig. 2.1.
2.2.1.2 Axial Forces
Axial forces due to short circuit current are produced by the radial magnetic flux. Axial
forces lead to tilting or bending of conductors and will be more dangerous when the
windings are not placed symmetrically. Any ampere-turn mismatch between LV and HV
windings will strengthen axial forces. Titling and bending of conductors between spacers
due to axial forces are shown in Fig. 2.2. Any small displacement during transformer
transportation or due to earth quake would result in intensified axial forces when short
circuit occurs. As the most common reported deformation type, symmetrical and
asymmetrical axial deformation in transformer winding will be discussed in detail in the
next Chapters.
15
(a) (b)
(c)
Figure 2.1. Winding deformation (buckling), (a) Free buckling (top view), (b) Forced buckling (top view), (c)
Free buckling (side view).
(a)
16
(b)
(c)
(d)
Figure 2.2. Winding deformation, (a) Before tilting, (b) After tilting, (c) Bending (side view), (d)Bending (close
view).
17
2.2.2 Transformer Transportation Causing Active Part Displacement
Transformer transportation is one of the major causes in transformer active part

displacement and winding deformation. Transformer must be carefully transported,
especially in the case of large power transformers. In this regard, there are four different
means to transport a transformer:
 Truck
 Railroad
 Sea carrier
 Air carrier
In each region, based on available transportation infrastructures one of the mentioned

carriers will be employed. From the safety point of view, the best approach should
produce the lowest vibration and bring high reliability even if it is slow and time
consuming. While a transformer is transported, the main tank is filled with dry air, dry
nitrogen or any other gas which does not have any chemical impact on transformer major
and minor insulation system and at the same time completely prevents the core and
windings from absorbing moisture until final on-site oil injection. Therefore, the pressure
of the gas inside the transformer tank must be higher than the outside. Also, bushings are
disassembled for easier carrying. A typical arrangement for transformer transportation is
shown in Fig. 2.3.
Figure 2.3. Transformer transportation schematic.
18
All mechanical forces which can have serious effect on the transformer tank during
transportation need to be considered when the transformer is settled on the portable
surface. Hence, in mechanical forces calculation, all of ramps, winding routes, wind force,
stop shock as well as bumps should to be taken into account. Indeed, all effective acting
forces must be less than the static friction force between the transformer tank and
portable surface plus stop forces. In this regard, the overall acting force can be calculated
as:
 Fradial   Fwind   Fpredicted   mg   Fstop

s (2.4)
where, Fradial is the imposed forces during transportation, Fwind is the wind force, Fpredicted
denotes other predictable forces, Fstop is the security force like bracing or stop forces, m is
the total transformer mass, and µs represents the static friction constant between the
transformer tank and portable surfaces and wooden bars if available. Figure 2.4 illustrates
different transformer transportation classes.
(a)
19
(b)
(c)
20
(d)
Figure 2.4. Transformer transportation, (a) Truck [61], (b) Rail road [62], (c) Sea carrier [63], (d) Air carrier
[64].
2.3 Winding Deformation Diagnosis Methods
Mechanical diagnostic methods have been developed to recognize transformer active part
displacement as well as winding deformation. Hence, various methods such as Low
Voltage Impulse (LVI), Frequency Response Analysis (FRA) and Short Circuit Impedance
(SCI) have been employed for off-line mechanical defects recognition in transformers [18].
In fact, off-line methods have been employed and their advantages from recognition
perspective have been debated quite extensively in the literature [18].
2.3.1 Short Circuit Impedance Measurement
The Short Circuit Impedance (SCI) may be considered as a parameter which highlights
imperfect magnetic coupling between primary and secondary windings. It contains
resistive and inductive terms, the latter being much more dominant than the former. The
short circuit impedance (or leakage inductance) can be represented as an additional
inductance in series with the transformer primary inductance, as shown in Fig. 2.5. A high
SCI value leads to a high voltage drop across the transformer terminals and thus affects
network voltage regulation, while a low value influences the network short circuit current.
21
The distances between the HV winding, the LV winding and the core of the transformer
have considerable influence on the SCI value (Fig. 2.6). The SCI is given by:
SCx Dx  f 
U x  0.248 2
K R   , (2.5)
2(Vn N w ) H m N B  50 
where Ux is the short circuit impedance, NB is the number of transformer core limbs
surrounded by HV and LV windings, KR is the Rogowsky coefficient taken as 1 for most HV
and LV winding arrangements [18] but which can be calculated if necessary, f is the
operational frequency, Hm=(Hm1+Hm2)/2, where Hm1 is the height of the LV winding and Hm2
is the height of the HV winding, Vn is the nominal voltage of the winding, Nw is the number
of winding turns and S is the apparent power quoted on the transformer nameplate. In
addition, Dx=DC2 + (BOW2-BOW1)/3 and Cx=C2 + (BOW1+BOW2)/3, where the distances C2, DC2,
BOW1 and BOW2 are shown in Fig. 2.6. Clearly, winding deformation (changes in the
geometrical factors) will result in a change of Ux.
Figure 2.5. Schematic model of primary, secondary and leakage inductances of a transformer.
The measured SCI for a transformer should be compared to the value printed on the
nameplate or quoted in factory test results. Winding displacement that may have
occurred since the factory tests were performed may then be detected. According to [65],
changes of more than ±3% should be considered as indicating winding deformation or
core displacement. IEC standard 60076-5 states that, changes should not exceed ±1% for
transformers with power rating capacities above 100 MVA [66].
SCI measurements can be performed on single or three-phase transformers, usually on the

HV winding, with the LV winding short-circuited. The cross-sectional area of the cable
used to short-circuit the LV winding must be at least 30% greater than that of the winding
conductor ‎[67], and it must be as short as possible. The resistance of the connection
between the LV terminals and the shorting cable must also be as small as possible. The SCI
measurement test setup for single and three-phase transformers is shown in Figures
2.7(a) and 2.7(b) respectively.
22
Figure 2.6. Schematic of transformer core and windings.
(a)
(b)
Figure 2.7. Short circuit impedance measurement setup, (a) Single phase transformer, (b) Three-phase
transformer. The test is made by short-circuiting the line-leads of the low-voltage windings and applying a
single-phase voltage at rated frequency to terminals of the other winding. Three successive readings are taken
on the three pairs of leads [65]. If the neutral terminal is available, the measurement can be conducted through
the line-lead and the neutral-lead.
23
2.3.2 Transfer Function Methods (FRA/LVI)
Transfer function is basically a way of describing a system behaviour. It is an established

convention for quantifying the system response as well as other means such as differential
equations. Transfer functions encapsulate information about frequency response and time
response given certain input signals, and system characteristics can be picked out through
the transfer function as well. There are two popular methods that can be used for transfer
function measurement (Fig. 2.8). The first one applies in the time domain and named as
Low Voltage Impulse (LVI), while the second one operates in the frequency domain and
called Frequency Response Analysis (FRA).
Figure 2.8. Transfer function measurement techniques.
Frequency domain measurement is performed by injecting a swept sinusoidal waveform

within a predetermined frequency band as it is described in (2.6).
U (t )  A sin(t )  A sin(2 f sweept ),

(2.6)
f min  f sweep  f max .
where A is the sinusoidal signal amplitude and fsweep denotes the variable frequency. It is
worth noting that for commercially-available FRA equipment, A varies from 5.66 Vpp to 25
Vpp for different manufacturers.
FRA data are commonly presented as magnitude Bode diagrams, with the x-axis for
frequency and the y-axis for the response magnitude. In some cases, more information is
also provided in the form of the FRA phase diagram, although this does not seem to
provide useful diagnosis on the mechanical integrity of transformer. In fact, the magnitude
diagram conveys more detail information than the phase diagram and that is the reason
most of literatures are just concentrated on the magnitude. Likewise, this study is focused
only on the magnitude response.
In general, FRA measurement is performed in the frequency band 20Hz – 1MHz for
transformers with highest voltage of > 72.5 kV, and in the range of 20Hz – 2MHz for
transformers with highest voltage of ≤ 72.5 kV ‎[1]. To be on the safe side, FRA measurement
can be performed over the range 20Hz – 2MHz for all transformers irrespective of their
voltage rating. However, in the case of special transformers or reactors, the upper limit may
24
be shifted to even higher frequencies. For instance with air-core reactors, this limit could be
increased up to 20 MHz. It should be noted that on-site FRA measurement beyond 2 MHz is
likely to experience undesirable oscillations or additional fluctuations.
Some researchers believe that reliable results are obtained only within the range 10 Hz - 1
MHz ‎[1] for routine transformers, while others have recommended an upper bound of 10
MHz ‎[73].
To facilitate classification, the frequency response data may be divided into three bands,
namely low-, medium-, and high-frequency bands (see Fig. 2.9). In [1] it is stated that the
data are dominated by the transformer core at low frequencies, by the winding structure at
medium frequencies, and by the connection leads at high frequencies. However, the
boundaries between the bands are not widely agreed.
Apart from the high-frequency band which is influenced by connections leads as well as
other setup contacts, the mid- and low- frequency bands of FRA spectrum should be
interpreted, precisely. To reach to this level, this study will argue that although the
transformer core affects mainly the low-frequency band, elimination of the transformer
core would totally removes its impacts on the FRA trace, and enable a more accurate
interpretation of the mid-frequency oscillations are become discussable. Thus, to conduct
precise modeling of transformer winding for mid-frequency interpretation, an air-core
model transformer was manufactured and studied in detail in Chapter 4. Chapter 3 also
provides the calculation of air-core transformer winding parameters to utilize in Chapter 4.
To conduct interpretation of low-frequency band oscillations, a full-size 100 kVA

transformer was used as the test object and low-frequency oscillations are discussed in
detail in Chapter 5.
25
Low Mid High
-10
-20
-30
-40
Magnitude [dB]
-50
-60
-70
-80
-90 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 2.9. Low, mid and high frequency bands of a typical FRA spectrum (measured on 400 MVA
transformer).
In the time domain method, the impulse waveform is injected into the test object input and
the time domain response is measured from the test object output. Once the time domain
measurement data is obtained, the transfer function in frequency domain can be
determined by using FFT technique.
Generally, the purpose of both methods is to excite the natural frequencies of the test
object. Technically speaking, the transfer function of a certain test object determined by
using frequency domain method is not identical to that calculated by employing time
domain measurement and then utilizing FFT technique. For consistency, researchers have
recommended applying only one method in the reiteration of the measurements on a
particular item of equipment over a certain period of time. LVI is generally faster than
FRA, while FRA is more accurate. The trend in industry shows more interest in the latter.
In transformer winding studies, the FRA could be measured for HV windings or LV

windings. In three phase transformers windings are inter-connected (eg. as wye, delta or
zigzag); the FRA data can be extracted collectively or independently for the windings.
In fact, the FRA can be measured for any proposed circuit when the transformer is out of
service. Definitely, if we are setting up a proposed circuit for transformer frequency
response measurement now, the same proposed circuit needs to be applied in future
studies of transformer frequency response. Otherwise the results would not be
comparable.
26
2.3.2.1 FRA Measurement Setups
FRA has been widely used as a comparative diagnosis method for some years now. The
initial FRA measurements during factory testing serve as the winding fingerprint
(reference, baseline, or original trace). Changes in winding configuration would almost
certainly cause changes in the frequency response trace ‎[17]. The initial measurements of
the distributed resistance, capacitance and inductance of a winding may also be usefully
compared with the same measurements following transformer maintenance, repair or
transport ‎[15], ‎[72].
Independent measurement of the frequency response for each individual winding is more
convenient for future comparison. Hence it is recommended for transformer frequency
response measurement the best simple and independent circuit to be considered.
Appropriate circuit for Yn winding connection will be phase bushing lead as input and
neutral bushing lead as output. When the transformer includes isolated wye or delta
connection, two phase leads need to be regarded as input and output. Based on this,
different test setups of FRA are given as follow ‎[1].
2.3.2.1.1 End-to-end Measurement
In this setup, FRA measurements can be made on a winding by injecting a preset signal Vin
at the line-lead, and detecting the response Vout at the neutral-lead of the transformer, as
shown in Fig. 2.10(a). The frequency response magnitude Kmag (the voltage attenuation in
dB) is given by:
V 
K  20log10  out  (2.7)
 Vin 
mag
2.3.2.1.2 Inductive Inter-winding Measurements
In this setup, FRA measurements can be made on a winding by injecting a preset signal Vin
at the line-lead (for instance HV winding, phase U), and detecting the response Vout at the
line-lead of the corresponding concentric winding (for instance LV winding, phase u),
where the neutral-lead of both windings are grounded as shown in Fig. 2.10(b).
2.3.2.1.3 Capacitive Inter-winding Measurements
In this case, FRA measurements can be made on a winding by injecting a preset signal Vin at
the line-lead (for instance HV winding, phase U), and detecting the response Vout at the line-
27
lead of the corresponding concentric winding (for instance LV winding, phase u), where the
neutral-lead of both windings are left open circuit as shown in Fig. 2.10(c).
2.3.2.1.4 End-to-end Short-circuit Measurements
In this case, FRA measurements can be made on a winding by injecting a preset signal Vin at
the line-lead, and detecting the response Vout at the neutral-lead of the transformer, where
all terminals of the other side are short-circuited as shown in Fig. 2.10(d).
Zin
Vsource
Vin
FRA
Vout
Transformer Zout
(a)
Zin
Vsource
Vin
FRA
Vout
Transformer Zout
(b)
28
Zin
Vsource
Vin
FRA
Vout
Transformer Zout
(c)
Zin
Vsource
Vin
FRA
Vout
Transformer Zout
(d)
Figure 2.10. FRA test setups, (a) End-to-end measurement, (b) Inductive inter-winding measurement, (c)
Capacitive inter-winding measurement, (d) End-to-end short-circuit measurement.
Figure 2.11 and Table 2.1 provide detailed FRA setups for both star and delta connections
(star and delta connections do not influence illustrated terminals in Table 2.1).
29
C1
B1
A1
A B C
HV Windings
A2
B2
C2
c1
b1
a1 a b c
LV Windings
a2
b2
c2
Figure 2.11. FRA test setups (detailed connections).
Table 2.1. FRA measurement connections (phases A, B and C).
Terminals Terminals connected

Setup Source (Vin) Response (Vout)
earthed together
End-to-end A1 A2 none none

Inductive inter-winding A1 a1 A2 and a2 none
Capacitive inter-winding A1 a1 none none
End-to-end short-circuit A1 A2 none a1-a2-b1-b2-c1-c2
End-to-end B1 B2 none none

Inductive inter-winding B1 b1 B2 and b2 none
Capacitive inter-winding B1 b1 none none
End-to-end short-circuit B1 B2 none a1-a2-b1-b2-c1-c2
End-to-end C1 C2 none none

Inductive inter-winding C1 c1 C2 and c2 none
Capacitive inter-winding C1 c1 none none
End-to-end short-circuit C1 C2 none a1-a2-b1-b2-c1-c2
30
2.3.2.2 Evaluation Indicators
A common method to interpret frequency responses of transformer windings is by using

evaluation indicators (statistical indices or statistical indicators). Such indicators have
been introduced by various researchers over the years. In order to exploit FRA test results
for extracting more information, a software package was developed to calculate statistical
indicators (see Appendix A). Statistical indicators with mathematical expressions as below
are implemented in the software to quantify the gap between reference and deviated
traces:
Correlation Coefficient (CC) [20]-[21], [23], [25]-[26]:
i1 X iYi
Ns
CC( X ,Y )  (2.8)
i1 X 2 i1Y 2
N s N s
i i
Maximum Absolute Difference (DABS) [20]:
i1| Yi  X i |
Ns
DABS( X ,Y )  (2.9)
Ns
Minimum Maximum (MM) [20]:
i1 min( X i ,Yi )

Ns
MM ( X ,Y )  (2.10)
i1 max( X i ,Yi )
N s
Standard Deviation (SD) [21]:
i1Y  X 2
Ns
SD( X ,Y ) 
i i
(2.11)
Ns 1
Spectrum Deviation (σ) [22], [68]:
2 2
  X i  Yi    X i  Yi 
 Xi      Yi   2 
 2     

1  (2.12)
Ns
 ( X ,Y ) 
Ns i 1   X i  Yi     X i  Yi  
  2     2  
       
31
where Xi and Yi are the ith elements of the fingerprint and measured FRA traces
respectively, and Ns is the number of elements (or samples). CC is thus a number whose
absolute value lies between 0 and 1 [21]. Maximum coefficient is achieved by complete
overlapping between the reference and measured traces and minimum coefficient is when
there is no overlapping at all.
SD values less than or equal to 1 indicate normal condition for windings, while values
greater than 1 reveal a problem associated with winding deformation [21]. The highest
possible similarity level of FRA traces can be determined by DABS, MM and σ parameters
once they take values as 0, 1 and 0, respectively but there are no predetermined limits for
maximum deviation for them. The Relative factor (Rxy) has been introduced and discussed
in [25] and [74]-[75] (see Appendix A). Tables 2.2 and 2.3 provide the criteria for CC, SD
and Rxy. The equation for Rxy is provided in Appendix A.
Table 2.2. CC and SD criteria[21].
Statistical indices Boundary values of parameters for deformation indication

CC (Magnitude spectrum) <0.9998
CC (Phase spectrum) <0.95
SD (Magnitude spectrum) >1.00
SD (Phase spectrum) >10.0
Table 2.3. Level of deformation based on Rxy [25], [74]-[75].
Deformation category Rxy

Severe RLF < 0.6
Obvious 1.0> RLF >= 0.6 or RMF < 0.6
Slight 2.0> RLF >=1.0 or 0.6 =< RMF < 1
Normal RLF >= 2.0, RMF >= 1.0 and RHF >= 0.6
LF: 1 kHz - 100 kHz , MF:100 kHz- 600 kHz, HF: 600kHz - 1 MHz
2.3.3 Deformation Coefficient Method
In this method which is recommended by [76], instead of sweep frequency, only three
measurements at selected high frequency and at selected low frequency are considered.
The main approach of this method is capacitance measurement at both ends of the
winding. Joshi et al [76] have introduced Deformation Coefficient (DC) as a judicable
parameter in transformer winding deformation recognition.
Transformer windings are generally represented by a lumped parameter equivalent
circuit which includes series capacitances and shunt capacitances. The Deformation
Coefficient is a function of changes in series capacitance as well as shunt capacitance.
32
Deformation Coefficient parameter which can be calculated through off-line measurement

is given by [76]:
 C  C 
DC  log10  1H 1H , (2.13)
 C  C 
 2H 2H 
where, C1H and C2H are the fingerprint values of measured terminal capacitances at the
selected high frequency and 1H and 2H are the terminal capacitance (after deformation)
values at terminals 1 and 2, respectively. These terminals are discussed in ‎[76] in detail.
As discussed in ‎[76], the proposed capacitances in (2.13) can be measured through off-line
process and DC deviation indicates transformer winding deformation. This method has
not been industrialized, but registered as a patent ‎[77].
2.4 FRA vs. SCI
Frequency response analysis and short circuit impedance measurement as two popular
methods for transformer winding deformation diagnosis will be employed to get insight
into transformer active part condition, but which one is more sensitive for checking
winding deformation? The answer to this question via a practical case study below
indicates FRA is a better technique.
A failed 400 MVA step-up transformer was used in order to compare the effectiveness of
FRA and SCI measurements. The principal parameters of the transformer are given in
Table 2.4. The transformer was suspected to have winding deformation due to the short
circuit fault. Therefore, both SCI and FRA measurement methods were performed to
assess the mechanical integrity of transformer and the results were compared.
2.4.1 SCI Measurements
The LV winding terminals were short-circuited, and SCI measurements were made
between the HV windings and the neutral terminal. The measured impedances are
presented in Table 2.5.
33
Table 2.4. 400 MVA step-up transformer parameters.
Parameter Value Parameter Value

Manufacture date May 1993 No-Load current (%) 0.37
Rated Voltage [kV] 242/20 Number of phases 3
Rated power [MVA] 400 Number of limbs 5
Rated current [A] 954/11550 Frequency [Hz] 50
No. of coolers 12 Cooling system OFAF
Table 2.5. Measured SCI values.
Zmeasured Zfactory Change

Measurement Scheme Measurement
[ohm] [ohm] (%)
HV-LV A-Neutral 17.028 17.290 1.515

HV-LV B-Neutral 17.798 17.290 2.938
HV-LV C-Neutral 17.088 17.290 1.168
Accuracy of multimeter: 0.5 mV
The percentage change, defined as 100 Zfactory – Zmeasured/ Zfactory, is smaller than the
maximum permitted change (3%) suggested in [65], for each of the three phases. Thus one
standard [65] indicates that winding deformation or displacement had not occurred.
(However, the change of 2.93% for phase B is close to the limit and thus might raise
concern to an expert). On the other hand, each of the three changes exceeds the 1% figure
as suggested in [66] for transformers with capacities above 100 MVA. Thus the latter
standard [66] indicates that deformation or displacement had occurred in each of the
three windings. Clearly the two standards yield conflicting indications.
2.4.2 FRA Measurements
FRA measurements were made by applying a 5.66 V signal across each winding, at 801
discrete frequencies in the range 20 Hz - 2 MHz. The characteristic impedance of the
measurement cables was 50 . The measured FRA traces were compared with fingerprint
traces obtained during transformer overhaul (before failure).
The results are shown in Fig. 2.12. In order to avoid compression of the higher frequency
data, the plots of the frequency response magnitudes are presented on a logarithmic
frequency scale. It can be seen that the measured and fingerprint spectra for phases A and
C are very nearly identical. However, this is not the case for phase B.
34
-10
-20
-30
-40
Magnitude [dB]
-50
-60
-70
Frequency Band: 20 Hz - 2 MHz

-80
Phase A (Measured trace)
Phase A (Fingerprint)
-90 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
(a)
-10
-20
-30
-40
Magnitude [dB]
-50
-60
-70

-80
Phase B (Fingerprint)
Phase B (Measured trace)
-90 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
(b)
35
-10
-20
-30
-40
Magnitude [dB]
-50
-60
-70

-80
Phase C (Measured trace)
Phase C (Fingerprint)
-90 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
(c)
Figure 2.12. (a), (b) and (c) show the measured and fingerprint frequency response magnitudes for phases
A, B and C respectively of the transformer HV side.
Low-, medium- and high-frequency bands for RXY, and deformation levels related to the RXY
values, have been defined in the Chinese standard [74] and by some other workers [25],
[75] (see Table 2.3). These definitions have been widely used for FRA trace evaluation, and
have been adopted in the present work. The calculated values of CC and SD for each of the
three phases are given in Table 2.6, and the corresponding values of RXY in Table 2.7. It can
be seen that, for phase B, CC and SD indicate deformation, and RXY indicates slight
deformation at low frequency and no deformation at medium and high frequencies. There
is no indication of deformation of phase A or phase C.
Table 2.6. CC, SD values
Frequency Band CC SD
Phase A Phase B Phase C Phase A Phase B Phase C
20 Hz - 2 MHz 0.9999 0.9977 1 0.5859 3.6367 0.5004
Table 2.7. Rxy values
Frequency Band RXY

Phase A Phase B Phase C
Low 10 1.2674 10
Medium 10 1.0724 10
High 10 2.3831 10
Low: 1 kHz - 100 kHz, Medium:100 kHz - 600 kHz, High: 600 kHz - 1 MHz.
36
These findings were checked and validated by internal inspection after de-tanking the
transformer. Figure 2.13 shows the side and front views of phase B of the HV winding.
Clearly the winding had suffered slight outward deformation (buckling) in two different
parts.
(a)
(b)
Figure 2.13. Buckled HV winding of phase B, (a) Side view of the middle disks, (b) Front view of the upper
disks.
37
2.4.3 Discussion on FRA and SCI
SCI measurements have been used for many years to detect and locate transformer
winding deformation and displacement. Given (2.5), the SCI values would be expected to
change, due to changes of Bow1 and Bow2, if significant radial movement occurred within a
winding. Changes in Hm would also be expected to influence the SCI value. However,
internal axial movement between transformer winding disks, e.g., bending and tilting of
conductors, would not be expected to influence the SCI value, since they would not change
any of the factors in (2.5). The results presented in this part suggest that FRA is capable of
providing reliable information on the level of deformation of transformer windings, based
on single values of the statistical quantities CC, SD and Rxy. However, it does not at present
provide information on the type of deformation. A more precise interpretation of FRA data
needs to be developed. This interpretation is provided in Chapters 4, 5 and 6.
2.4.4 Summary on FRA and SCI Methods
Off-line SCI and FRA measurements were conducted on a transformer which had failed
because of deformation of the B phase of the HV winding. Interpreted according to IEEE
Standard 62-1995 [65], the SCI values indicated that winding deformation had not
occurred, but when interpreted according to IEC Standard 60076-5, Ed. 3.0, 2006 [66],
they indicated that deformation had occurred in each of the three windings. Two indices
derived from comparison of fingerprint and measured FRA data indicated deformation of
phase B winding, and the third index indicated slight deformation of that winding. Visual
inspection of the failed transformer showed that the B phase winding had indeed been
deformed.
Since SCI measurements are made only at the operating frequency, it is expected that they
would provide less detailed information on the state of the windings than the more
comprehensive FRA measurements. Nevertheless, it is recommended that both types of
measurement be performed on power transformers (along with other conventional test
measurements) as part of transformer routine maintenance.
2.5 Conclusion
This Chapter provided a review of the main reasons causing winding deformation and
specifically discussed the impact of short circuit current. The various methods to
recognize winding deformation and active part displacement were introduced. It was
highlighted that FRA is superior to SCI in terms of winding deformation recognition.
Statistical indicators as the evaluation methods of frequency response were discussed.
38
Chapter 3. Transformer Winding Parameters
Chapter 3
Transformer Winding Parameters
3.1 Introduction
In order to gain new knowledge on FRA trace interpretation through analytical approach,
the focus should be specifically on the transformer winding electrical parameters. To this
end, the present Chapter investigates inductance and capacitance calculation in
transformer windings. This in turn will help in verification of the developed model in the
next Chapter. It is also quite crucial to study axial and radial deformation in transformer
winding in Chapter 6.
To conduct the calculations, the test object is a transformer winding. This is just a model to
clarify how to calculate inductance and capacitance. The model winding is enclosed in a
cylindrical metal container (tank). At first, the self- and mutual-inductances as well as
series and shunt capacitances are calculated for the winding in normal (not deformed)
condition. Afterwards, the analytical approach is validated through practical measurement
on the manufactured small model transformer with air-core continuous disk type HV and
LV windings. This circumstance verifies that the calculated results are accurate.
It should be highlighted that the problem of inductance calculation has been quite
thoroughly treated in a number of significant studies. Formulas for circular filaments were
first given by Maxwell ‎[78]. Subsequently, Rayleigh ‎[79], Lyle ‎[80], Butterworth ‎[81], Snow
‎[82], Rosa ‎[83], Curtis and Sparks ‎[84], Grover ‎[85], Babic ‎[86], Conway ‎[87], etc., have
developed and derived others formulas and tables to calculate self and mutual
inductances.
In this study, the Grover’s formulas ‎[85] are used for the self-inductance calculation of
circular coils of rectangular cross section as well as mutual-inductance calculation for
coaxial circular filaments, circular filaments whose axes are inclined and circular coils
with parallel axes. Grover’s formulas are then developed in Chapter 6 for the calculation of
transformer winding inductance matrix for the case of axial deformation.
39
It should be noted that to simplify inductance calculations, the thickness of paper

insulation wrapped over the conductors is ignored in this study, but for capacitance
calculation this thickness is taken into consideration.
3.2 Self and Mutual Inductances of Transformer Winding (Analytical

Approach)
In the absence of magnetic materials, self and mutual inductances are parameters
dependent only on the system configuration and independent of the current ‎[85].
Formulas for the self-inductance calculation of the disk winding have been discussed
widely in the literatures ‎[88]-‎[89]. However, the calculation of mutual-inductance between
disks has not been formulated ‎[88]. The inductance matrix of the winding (Leq) illustrated
in Fig. 3.1 is given by (3.1) in general form and (3.2) in detailed form:
 L M AB M AC M AD 
 A

M LB M BC M BD 
Leq   BA  (3.1)
M M CB LC M CD 
 CA 
M M DB M DC LD 
 DA
 L1 M12 M13 M14 M15 M16 M17 M18 M19 M110 M111 M112 M113 M114 M115 M116 
. L M 23 M 24 M 25 M 26 M 27 M 28 M 29 M 210 M 211 M 212 M 213 M 214 M 215 M 216 
 2
. L3 M 34 M 35 M 36 M 37 M 38 M 39 M 310 M 311 M 312 M 313 M 314 M 315 M 316 
 
. L4 M 45 M 46 M 47 M 48 M 49 M 410 M 411 M 412 M 413 M 414 M 415 M 416 
. L5 M 56 M 57 M 58 M 59 M 510 M 511 M 512 M 513 M 514 M 515 M 516 
 
. . L6 M 67 M 68 M 69 M 610 M 611 M 612 M 613 M 614 M 615 M 616 
. . L7 M 78 M 79 M 710 M 711 M 712 M 713 M 714 M 715 M 716 
 
. . L8 M 89 M 810 M 811 M 812 M 813 M 814 M 815 M 816 
Leq    (3.2)
. . . L9 M 910 M 911 M 912 M 913 M 914 M 915 M 916 
. . . L10 M1011 M1012 M1013 M1014 M1015 M1016 
 
 . . . L11 M1112 M1113 M1114 M1115 M1116 
. . . . L12 M1213 M1214 M1215 M1216 
 
 . . . . L13 M1314 M1315 M1316 
 
. . . . L14 M1415 M1416 
. L15 M1516 
 
 . . . . . . . . . . . . . . . L16 
where:
 L1 M12 M13 M14 

M L2 M 23 M 24 
LA    (3.3)
21
 M 31 M 32 L3 M 34 
 
 M 41 M 42 M 43 L4 
40
 L5 M 56 M 57 M 58 
M L6 M 67 M 68 
(3.4)
LB   
65
 M 75 M 76 L7 M 78 
 
 M 85 M 86 M 87 L8 
 L9 M 910 M 911 M 912 

M L10 M1011 M1012 
LC   
109
(3.5)
 M119 M1110 L11 M 1112 
 
 M129 M1210 M1211 L12 
 L13 M1314 M1315 M1316 

M L14 M 1415 M 1416 
LD   
1413
(3.6)
 M1513 M1514 L15 M 1516 
 
 M1613 M1614 M1615 L16 
 M15 M16 M17 M18 

M M 26 M 27 M 28 
 
25
M AB (3.7)
 M 35 M 36 M 37 M 38 
 
 M 45 M 46 M 47 M 48 
 M19 M110 M111 M112 

M M 210 M 211 M 212 
  (3.8)
29
M AC
 M 39 M 310 M 311 M 312 
 
 M 49 M 410 M 411 M 412 
41
 M113 M114 M115 M116 

M M 214 M 215 M 216 
 
213
M AD (3.9)
 M 313 M 314 M 315 M 316 
 
 M 413 M 414 M 415 M 416 
 M 59 M 510 M 511 M 512 

M M 610 M 611 M 612 
 
69
M BC (3.10)
 M 79 M 710 M 711 M 712 
 
 M 89 M 810 M 811 M 812 
 M 513 M 514 M 515 M 516 

M M 614 M 615 M 616 
 
613
M BD (3.11)
 M 713 M 714 M 715 M 716 
 
 M 813 M 814 M 815 M 816 
 M 913 M 914 M 915 M 916 

M M1015 M 1016 
 1013 M 1014 
M CD  (3.12)
 M1113 M1114 M1115 M1116 
 
 M1213 M1214 M1215 M1216 
In the above, LA is the inductance matrix of disk A, MAB is the mutual-inductance matrix of
disks A and B, L1 denotes the turn inductance, and M12 is the turn-to-turn inductance
between the turns 1 and 2. Kirchhoff ‎[90] has shown that the equivalent self-inductance of
a disk is equal to the summation of the self and the mutual-inductance of each turn with
respect to all the other turns in that disk. Hence, for the proposed winding, the self-
inductance of the first disk (LA) is given by:

LA  L1  L2  L3  L4  2 M12  M13  M 23  M14  M 24  M 34  (3.13)
while the mutual-inductance of the same disk is given by:
42

M A  2 M12  M13  M 23  M14  M 24  M 34  (3.14)
Equations (3.13) and (3.14) can be extended for any number of disks and not restricted to
a single disk; hence, the self-inductance for the winding of Fig. 3.1 is given by:

Leq  LA  LB  LC  LD  2 M AB  M AC  M BC  M AD  M BD  M CD 
(3.15)
The mutual-inductance of the winding is then defined as:

M eq  2 M AB  M AC  M BC  M AD  M BD  M CD  (3.16)
M13 M24 M34 M23 M12

Air-Core
εt
1 22 33 44 4 3 2 1
A
1 4 3 2 1
MAB δt
8 7 6 5
5 5 6 7 8
B
8 7 6 5 6 7 8 MAC
MBC d MAD
9
9 10
10 11
11 12
12 12
12 11
11 10
10 99 C
MBD
MCD δd
16 15 14 13 13 14 15
D
16 15 14 13 13 14 15 h
B w
W R
Figure 3.1. Air-core transformer winding model.
It should be noted that (3.2) is a symmetrical matrix for a winding with normal structural
condition. It will be shown latter that this matrix will turn into an asymmetrical matrix
once axial deformation occurred.
3.2.1 Self - Inductance
The self-inductance of a circular winding with rectangular cross sections is a function of

winding shape ‎[85]. The significant parameters defining the winding shape are the mean
radius of the turns per disks, and axial and radial dimensions of the conductor cross
43
sections. These parameters are illustrated in Fig. 3.1. Based on this, the self-inductance (L)
of an air-core circular disk for h=W is given by ‎[85]:
L  0.001RN 2P0 H (3.17)
where, N is the number of disk turns, R is the mean radius of disk turns, h is the axial
dimension of the conductor cross section, W is the radial dimension of the winding cross
section, and P0 is a function of W/2R. For relatively small cross section category such that
(W/2R < 0.2), P0 is given by ‎[85]:
  2  2 2

P0  4   1  1  W 
  ln  8 2R    0.84834  0.2041 W  
2 12 2 R    W      (3.18)
     2R  
    
For a thin circular disk with rectangular cross section of any desired proportions, the self-
inductance is given by:
 
L  0.019739  2R  N 2R( K  k ) H (3.19)
W  N 0
where KN comes from Nagaoka’s formula ‎[91] and can be derived through Table B.1 in the
[Appendix B], and k0 is a factor that specifies circular inductance decrement due to the
separation of turns in radial direction. For a single turn with significant mean radius
dimension, k0 can be zero (see Table B.2).
3.2.2 Mutual – Inductance
The mutual inductance of a transformer winding (Mc) can be calculated for one disk with
respect to another disk as given by (3.20), or it can be achieved through summation of the
turn-to-turn mutual inductances in the inductance matrix of a winding as obtained by
(3.21).
Mc  N1N2M0 H (3.20)
N1 and N2 denote the numbers of turns of different disks, and M0 is calculated using Lyle’s
method ‎[80].
In this study, as the main goal is specifically focused on the axial and radial deformation in
a disk; thus, instead of winding mutual-inductance in (3.20), the turn-to-turn mutual-
inductance calculation is further discussed in detail. Indeed, winding mutual-inductance is
44
simply achievable through turn-to-turn mutual-inductance summation. The mutual-

inductance of coaxial circular filaments (turn-to-turn) is given by ‎[85]:
Mf
m
Rb Ra H (3.21)
where Ra and Rb are the mean radius of the turns a and b, respectively (see Fig. 3.2(a)). fm
is a function of parameter k which is given by:
k  (1  )2   2 ,
2 2
  Ra ,  d (3.22)
(1  )   Rb Rb
d is the distance between circular turns as illustrated in Figures 3.2(a) and 3.2(b).
Ra
Rb
(a)
Ra
Rb
(b)
Figure 3.2. Coaxial circular conductors, (a) Coaxial filaments, (b) Coaxial disks.
For usual transformer windings, k ≤ 0.1; thus, fm can be obtained by ‎[85]:
 
f  0.014468 log10 1  0.53307  (3.23)
m  k 
For the mutual-inductance between two turns in a common disk β=0, and α=1 is
considered for the inter-disk mutual inductances of the conductors with equal mean
radius. Therefore, all mutual inductances (Mxy) in (3.2) can be obtained through the use of
(3.21).
45
3.3 Series and Shunt Capacitances of Transformer Winding
3.3.1 Series Capacitance
3.3.1.1 Layer Winding
Figure 3.3 is a schematic of layer winding which shows its series and shunt capacitances.
According to Fig. 3.3, there is a capacitance between each conductor and its adjacent
conductors. If Nw is the number of turns in layer winding, the number of series
capacitances would be Nw-1, and the series capacitances can be determined through ‎[92]:
Ctt (3.24)
Cs 
Nw 1
where Ctt is the turn-to-turn capacitance and Cs is the total series capacitance in a layer
winding. It should be noted that the value of series capacitances will be reduced when the
number of series conductors is increased.
U
Air-Core Tank
Cg
1
Ctt
Cg
2
2
Ctt
Cg
33
Ctt
Cg
44
Next turn
Figure 3.3. The overall layout of a layer winding including equivalent capacitance network.
3.3.1.2 Disk Winding
Various types of disk winding are used in power and distribution transformers.
Continuous disk winding (also called conventional disk winding) and interleaved winding
46
are the most popular. In some distribution and power transformers continuous disk
winding is used, while interleaved winding is employed in large power transformers at
voltage level of 230 kV and above.
3.3.1.2.1 Continuous Disk Winding and Related Series Capacitances
In Fig. 3.4, a continuous disk winding containing two disks is shown. Also, Fig. 3.5 shows
the equivalent capacitive network of continuous disk winding. The series capacitance is
divided into two parts:
1) Total series capacitance between the turns, Ct.
2) Total series capacitance between the disks, Cd.
To calculate the equivalent series capacitance in continuous disk winding, the energy
summation method is used. According to this method, the summation of energies in the
capacitances along a pair of disks is equal to the total energy which exists in the winding
with those two disks. It is assumed that the number of conductor turns in each disk is N.
The number of series capacitors between turns, as shown in Fig. 3.5, will be 2N-2 for a pair
of disks.
Therefore the total equivalent capacitance between the conductors, Ct, is given by ‎[92]:
2
1 1  U  1  N 1 
CtU 2  (2 N  2) Ctt    Ct  Ctt   (3.25)
2 2  2N  2  N2 
Air-Core Line-lead Tank
4
4 3
3 2
2 1
1
5
5 6
6 7
7 8
8
Next disc
Figure 3.4 Continuous disk winding schematic, taken and modified ‎[92].
47
Line-lead
Air-Core Tank
Ctt Ctt Ctt
4
4 3
3 2
2 1
1
Cdd Cdd Cdd U
5
5 6
6 7
7 8
8
Next disc
Figure 3.5. Equivalent capacitance network of the continuous disk winding.
where U is the voltage drop across the pair of disks (see Fig. 3.5), and Ctt is given by:
 h  2 t 
Ctt   t  0 R   . (3.26)
 2 t 
In the above formula, δt is the thickness of inter-turn insulation, εt is the relative

permittivity of paper insulation, and ε0 is the vacuum permittivity.
Calculation of the equivalent capacitance between disks (Cd) is based on the voltage
distribution demonstrated in Fig. 3.6.
According to Fig. 3.6, when moving from end points starting from conductor number 1 or
number 8 towards middle of the winding (conductor number 4 and number 5) the voltage
on corresponding conductors will change linearly and continually. Hence, the steady state
voltage distributions for conductor in upper and lower disks are as follows:
2   x x
U up (n)  U , U down (n)  U (3.27)
2  2 
where n is the turn number, λ' is the total length of conductor in one disk, and U denotes
the voltage across the disk pair. The equivalent inter-disk capacitance between two disks
is given by ‎[92]:
1 1
Ed  CdU 2  Cdd (U up (n)  U down (n))2
2 2

(3.28)
1 x
 CddU 2  (1  ) 2 dx
2 0 
Cdd
 Cd   . (3.29)
3
48
l
Ctt
1 2 3 4
U Cdd
8 7 6 5
U
U/2N
U/2N
x
0
l
Figure 3.6. Pair of disks, cross-section overview and voltage distribution along disks pair (paper insulation has
been ignored).
The summation of Ct and Cd gives the equivalent series capacitance Cs-pair for a pair of disks
in a continuous disk winding:
Cs pair  Cd  Ct . (3.30)
The first part in (3.30) corresponds to the capacitance between the disks, which is
obtained with stored energy and the second part corresponds to the capacitance between
the conductors. The equivalent capacitance for the entire winding (Cs) is then obtained by:
 N  
 
  w  1

N 
1  Nd  1 
Cs  4 Cd   d  2 Ctt  (3.31)
Nd  Nd N  
  w 
 N  
  d  
where Nd is the number of transformer winding disks, and Nw is the number of winding
turns.
3.3.1.2.2 Interleaved Winding and Related Series Capacitances
In an interleaved winding, Cs increases considerably as compared to continuous disk

winding, therefore it is utilized to improve the electric stress distribution. Various
methods to interleave the disk windings are available. One of the simplest techniques is
shown in Fig. 3.7.
49
4
6 3
2 2
5 1
1
5
3 6
7 7
4 8
8
Next disc
Figure 3.7. The interleaved disk winding.
The series capacitance of this interleaved winding is given by [92]:
C ( N  1)
Cs  Eint tt (3.32)
4
where Eint is the number of disks used for interleaving and equal 2 for Fig. 3.7. Increasing
Eint will increase the manual welding work required within the interleaved winding
considerably. The magnitude of series capacitance is increased in interleaved windings
and this leads to a more uniform voltage distribution throughout the winding. However,
the turn-to-turn potential difference is increased considerably for steady state operation
and thus the turn-to-turn insulation should be made properly to withstand against this
potential.
3.3.1.2.3 Intershield Winding
In Fig. 3.8, the configuration of a disk winding with electrostatic shields in each disk is
shown. In this winding, the shield turns, which can be made from copper or aluminum
conductor, are placed between the winding main conductors at predetermined places,
while the shield or shield turns of each disk are insulated from the conductors [Appendix
C]. Electrostatic shield conductors from the upper disk (in a pair of disks) are connected to
the electrostatic shield conductors of the lower disk at outermost shield turn as shown in
Fig. 3.8. For instance, the shield conductor between main conductors 1 & 2 of the upper
disk is connected to the shield conductor between conductors 7 & 8 from the lower disk
and both are isolated from the main conductors.
50
4
4 1 3
3 2 2
2 3 1
1
4
5 6 3
6 5 2
7 4 1
8
Next disc
Figure 3.8. The intershield disk winding.
The series capacitance of the intershield winding is given by [Appendix C]:
1  N  1 2 N sh ( N  1)  Cdd
Cs  Ctt     (3.33)
 N 
2
2 N 3
where Nsh is the number of shield turns per disk.
3.3.2 Shunt Capacitance
The shunt capacitance between the winding and the cylindrical metal container (tank) is
given by:
2 H w
Cg  (3.34)
ln  r2 r1 
where, Hw is the height of the winding, ε is the dielectric permittivity, r1 is the radial
dimension of the winding, r2 denotes the radial dimension of the tank, and δt as compared
to r2 is small and thus ignored.
3.4 Verification of Calculated Parameters Using Manufactured Model

Transformer
3.4.1 Manufactured Model Transformer (Test Object)
In order to conduct this thesis, a specific test object was designed and fabricated by the
author. This test object is a model transformer with air-core concentric continuous disk
type HV and LV windings. The HV winding consists of 8 disks with 8 conductor turns per
51
disk. The LV winding has 10 disks with 6 conductor turns per disk. The cylindrical tank
housing the windings was made from plexiglass material. The line and neutral leads of the
windings were brought out from the tank through appropriate HV and LV bushings. A
drain valve was installed on the top plate to enable oil injection and also taking oil sample.
The plexiglass tank was air-tight sealed, proper for vacuuming and winding dry-out. Also
when required, an aluminum foil was wrapped over the glass tank to simulate the metal
tank. Detailed information of the manufactured test object is provided in Appendix D. The
test object is shown in Fig.3.9.
(a)
(b)
Figure 3.9. Manufactured glassy model transformer (a) Bird’s-eye view, (b) Side view.
52
3.4.2 Inductance Calculation of the Test Object
The inductance of the HV winding of the test object was measured using a precision LCR
bridge (AIM-TTI LCR400) and compared with the value calculated based on the analytical
approach developed in this Chapter. The measurement value obtained is:
LHV (meas)  710.7000 H (3.35)
as shown on the LCR Bridge in Fig. 3.10.
Figure 3.10. Measured value for the inductance of HV winding.
For the inductance calculation using analytical approach, the HV winding schematic of the
manufactured test object as shown in Fig 3.11 was considered. Based on this and ignoring
the paper insulation thickness of the conductors, the calculated value for inductance is
given as (3.36):
 21.1170 17.0926 11.8346 8.7993 6.6873 5.0901 3.8249 2.7928 

17.0926 21.1170 17.0926 11.8346 8.7993 6.6873 5.0901 3.8249

 
11.8346 17.0926 21.1170 17.0926 11.8346 8.7993 6.6876 5.0901 
 

8.7993 11.9346 17.0926 21.1170 17.0926 11.8346 8.7993 6.6873 
LHV
6.6873 8.7993 11.8346 17.0926 21.1170 17.0926 11.8346 8.7993 
 
5.0901 6.6873 8.7993 11.8346 17.0926 21.1170 17.0926 11.8346 
3.8249 5.0901 6.6873 8.7993 11.8346 17.0926 21.1170 17.0926 
 
 2.7928 3.8249 5.0901 6.6873 8.7993 11.8346 17.0926 21.117 
(3.36)
Hence, the total value of the HV winding inductance (LHV) is given by:
53
LHV (calc)  743.2651 H (3.37)
Equation (3.37) in turn verifies the analytical approach. The discrepancy between
measured and calculated values is less than 10 percent and could be due to the paper
insulation thickness being ignored.
M13 M24 M34 M23 M12

Air-Core
εt
11 22 33 44 1
5 26 37 48
MAB δt
1
16 2
15 3
14 4
13 1
12 2
11 3
10 49
MBC d
1
17 2
18 3
19 4
20 1
21 2
22 3
23 4
24
MCD δd
1
32 2
31 3
30 4
29 1
28 2
27 3
26 4
25
MCE
MDE
MBH 1
33 2
34 3
35 4
36 1
37 2
38 3
39 4
40
MEF
MEG 1
48 2
47 3
46 4
45 1
44 2
43 3
42 4
41
MCG
1
49 2
50 3
51 4
52 1
53 2
54 3
55 4
56
MGH
1
64 2
63 3
62 4
61 1
60 2
59 3
58 4
57
B
w
R
W
Figure 3.11. HV winding schematic (glassy model transformer), R= 95 mm, w= 3mm, W= 24 mm d= 11mm, δd=
6 mm and conductor height h= 7 mm.
54
To verify the capacitance calculations, the HV winding shunt capacitance of the test object
was measured using Omicron Mtronix MI600 DDF measurement system and the result
was compared to analytical approach (see Fig. 3.12). This capacitance contains the
capacitance of HV winding with respect to LV winding including pressboard and the
capacitance between HV winding and the aluminum tank. The measured and calculated
results are obtained as (3.38) and (3.39), respectively. Figure 3.13 shows the top view of
the model transformer schematic.
Cg  98.2700 pF (3.38)
HV ( meas )
Cg  87.2861 pF (3.39)
HV ( calc )
Figure 3.12. Measured value for shunt capacitance, HV winding.
24
mm
8m
m Plexiglass Tank
18
mm HV
2m
m Pressboard
LV
m
m
55
95 mm
11
6
m
m
Figure 3.13. Model transformer schematic, top view.
55
The measured value for the HV shunt capacitance shows a larger value than the calculated
one. This difference is not quite significant particularly when the measurement is
performed in the range of (pF). Nevertheless, the reason should be taken into
consideration. In fact, this discrepancy comes through the impact of the capacitance from
the bushing and the aluminum tank. Physically speaking, the bushing capacitance is in
parallel with the other calculated capacitances and in turn increases the total shunt
capacitance of the HV winding in practice.
3.5 Conclusion
This Chapter has derived the self and mutual inductances between the transformer
winding turns. The turn-to-turn inductance relationships were used to develop the
inductance matrix and derive the final self and mutual inductances of the winding. The
series and shunt capacitances of transformer winding were then derived through a similar
approach. A small air-core glassy model transformer was fabricated for experiment and
analytical calculations were verified against practical measurement results on this test
object. The inductance matrix and capacitance values are essential components in the case
of winding model verification as well as axial and radial deformation studies in Chapters 4
and 6.
56
Chapter 4. Transfer Function Model of Air-Core Transformer Winding
Chapter 4
Transfer Function Model of Air-Core

Transformer Winding
4.1 Introduction
FRA data are typically reported as Bode diagram over a determined frequency band. It was
discussed earlier that FRA data in the mid-frequency band are influenced by the winding
structure, while data in the low-frequency band are affected mainly by the transformer core.
To study just the impact from winding structure on the FRA spectrum, this Chapter has
concentrated on the mid-frequency oscillations. Mathematical approach using travelling
wave theory is employed to explore frequency response trace behaviour. Practical studies
on a small air-core model transformer as well as two 66 kV, 25 MVA continuous and
interleaved disk windings have been performed to validate the mathematical calculations. In
addition, two 245 kV, 45 MVA and 66 MVA power transformers are tested to obtain mid-
frequency oscillations and compare these results with mathematical evaluation.
4.2 Modelling
Significant studies have concentrated on wave propagation in transformer windings [93]-

[98]. Loss-less charging and discharging currents and also wave propagation in transformer
winding using Maxwell’s equations for nth turn have been calculated in [94]. In addition,
travelling wave and multi-conductor transmission line theories have been utilized in [95] to
calculate voltage and current propagation for transient studies. In [95], it was emphasized
that the oscillation characteristics in the transformer primary winding are essentially the
same as that obtained when the secondary winding is ignored. Hence, it is appropriate to
perform the analysis for a single independent winding. The equations then become greatly
simplified and easy to visualize [95].
In this Chapter, the above mentioned fundamental studies as well as calculations are utilized
and extended in detail to interpret FRA mid-frequency band. To this end, the equivalent
57
detailed circuit of transformer winding as shown in Fig. 4.1 is considered. According to this
model, the charging currents for an infinitesimal length of transformer winding are given by:
 2e e
ics  cs , ig  g (4.1)
xt x
(a)
(b)
Figure 4.1. Equivalent detail circuit of a transformer winding (dx denotes an infinitesimal length of winding), (a)
Entire winding schematic [95], (b) Close view of the conductors and modeled parameters.
where, ics is the internal capacitive current per unit conductor length, cs denotes series
capacitance, e is the potential to ground (time domain), ig represents external conductance
current per unit conductor length and g is the shunt turn-to-turn conductance.
Also, discharging currents through shunt capacitance as well as conductance for
infinitesimal length of transformer winding flow to ground are calculated as equation (4.2):
e
ic  cg , iG  eG (4.2)
g
t
58
where, icg is the external capacitive current per unit conductor length, cg denotes the shunt
capacitance to ground, iG is the external conductance current per unit conductor length, and
G represents the shunt conductance to ground. Equation (4.3) is valid if the transformer
winding includes numerous number of turns and also if the detailed distribution of the wave
propagation along a single turn length is not of interest [94]. Therefore, the space derivative
of the current is calculated as equation (4.3) [95]:

(i  ics  ig )  icg  iG  0, (4.3)
x
where, i is the conductor current. Substituting of equations (4.1) and (4.2) into equation
(4.3) will lead to equation (4.4):
  2e e e
(cs  g  i)  cg  eG  0 (4.4)
x xt x t
According to (4.4), the space derivative of the winding current for an infinitesimal length of
transformer winding is given by:
i 3e  2e e
 cs 2  g 2  cg  eG (4.5)
x x t x t
On the other hand, the self-inductance of an infinitesimal winding turn contributes an

induced voltage. In addition, two adjacent turns coupled by the mutual-inductance cause
additional induced voltage in the turn.
Therefore, the space derivative of winding voltage including consideration of conductor loss
for an infinitesimal length of transformer winding can be calculated as equation (4.6) while
the partial mutual inductances due to other conductors’ turns are neglected:

e
x
    ti  ri
  (4.6)
where:
      l (4.7)
59
γ and μ are the self- and mutual-inductance per unit conductor length of the turn and
adjacent turn, respectively, r is the conductor resistance, and l is the inductance coefficient.
Subsequently, the voltage difference along the coil can be calculated as follow:
e i
 l  ri (4.8)
x t
In fact, the voltage as well as the current in a long coil varies from turn-to-turn as a function
of time and also space.
In [42], current and voltage studies have been converted from time domain to frequency
domain using Laplace transform while initial conditions have been supposed to be zero.
Therefore, transferring equations (4.5) and (4.8) from time domain into frequency domain
yield:
jl  r
 V ( j , x)       I ( j , x) 
0

j c G  V ( j, x)  

x  I ( j, x)   0
   g  
(4.9)
0 0   2  I ( j , x) 
   
 0 ( jcs  g )  x2 V ( j, x) 
 
where, V represents potential to ground (frequency domain), I is the conductor current

(frequency domain), ω is the angular frequency, and j denotes the imaginary operator.
According to equation (4.9), the differential equation for voltage propagation along
transformer winding is as follows:
 2V ( j, x)  2V ( j, x)
  ( jcg  G)V ( j, x)  ( jcs  g ) (4.10)
( jl  r )x x 2
 2V ( j, x) ( jl  r )( jcg  G)

 V ( j, x)  0. (4.11)
x 2 (1  ( jl  r )( jcs  g )
Therefore, the solution for the voltage at any point x along the transformer winding is given
by:
V ( j, x)  A1e x  A2e x (4.12)
60
( jl  r )( jcg  G )
 (4.13)
1  ( jl  r )( jcs  g )
β denotes the propagation coefficient and A1 and A2 are constants that depend upon the
boundary conditions. It is worth noting that x=0 and x=λ correspond to the neutral- and line-
terminal points of the transformer winding, respectively where the winding length is λ.
According to equation (4.9), the current at any point of the transformer winding is
calculated as equation (4.14):
I ( j , x) 

jl  r

A1e  x  A2e  x .  (4.14)
The frequency response magnitude for a transformer winding is calculated as equation

(4.15). Fig. 4.2 demonstrates common FRA test setups for a single winding and also
transformer including the source voltage as well as Vin and Vout. Based on Fig. 4.2, the output
voltage with respect to input voltage should be calculated to reach frequency response. This
calculation was given in (2.7).
(a) (b)
Figure 4.2. Common FRA test setup, (a) For single winding (b) For transformer.
According to equation (2.7) and the FRA test setup shown in Fig. 4.2; Vin is measured at the
line-terminal of the winding while Vout is measured at the neutral-terminal. Also, Zin and Zout
are the input and output measurement cable impedances and commonly represented by 50
Ω resistors. Since, Zout has small value compared to equivalent impedance of the winding, it is
approximated that Vout << Vin. Therefore in equation (4.12), A1 = - A2 and the frequency
response trace is obtained as:
61
Vout Z out I ( j , 0) Zout   e  0  e  0 

    (4.15)
Vin V ( j ,  ) jl  r  e   e 
It should be emphasized that according to equation (4.15) and above mentioned

assumption, the transformer frequency response magnitude exhibits small value especially
in very high frequencies. Although, according to equation (2.7) when equation (4.15) is
subject to logarithm operation, reported frequency response magnitude in Bode diagram
would be significant and completely traceable. Simplification of (4.15) using complex
algebra is given by:
Vout  1 
  (4.16)
Vin  j sin( j  ) 
Zout  ( jl  r )1   ,   Nwd  (4.17)
where, Nw is the number of winding turns, and dʺ is the average length of a conductor turn.
Transfer function resonances occur when the denominator in equation (4.16) experiences
minimum absolute value. It can be yielded once β2λ2=-(krπ) 2. Therefore:
 ( jl  r )( jcg  G )  2
    kr 
2 2

 1  ( jl  r )( jcs  g )  (4.18)
kr  1,2,3,.......
then:
 2 (lcg  kr2 2 ( 2 ) 1lcs ) 

 ((rcg  lG ) j  kr2 2 ( 2 )1(rcs  lg ) j )  (4.19)
rG  kr2 2 ( 2 )1 rg  kr2 2 ( 2 ) 1
Moving from low frequencies to mid and ultimately high frequencies will lead to ω2 >> ω,
therefore:
 2 (lcg  kr2 2 ( 2 )1lcs )  kr2 2 ( 2 )1 (4.20)
Hence, the resonance frequency (fr) will be calculated as (4.21):
62
 
kr  1 
fr  (4.21)
2  (c  c k 2 2 ( 2 )1)l 
 g s 
4.3 Discussion on Resonant Frequencies
Technical specifications and physical characteristics for each and every transformer are
major factors in determining the limits of the low-, mid- and high-frequency bands. The most
appropriate limits for each are still not widely agreed. Nevertheless, the flux division theory
and short circuit principles ‎[17] can be used to determine the low-frequency band. High-
frequency band can be recognized through test setup variation and then the remaining gap
constitutes the mid-frequency band.
Regarding the low-frequency band, as the inductive reactance of the winding is greater than
the capacitive reactance; low frequency behaviour is affected mainly by transformer
inductive reactance and the Bode diagram experiences a falling trend. Furthermore, the self-
inductance of transformer winding is greater than mutual inductances. Hence, low
frequency behaviour is affected mainly by transformer inductive reactance due to the
winding self-inductance.
Mainly comprising series and shunt capacitances, the capacitive reactance in the case of
high-frequency band could be neglected as its reactance is reaching sufficiently low value.
Hence, the measurement cables as well as connection resistances of the test setup would
become dominant at the high-frequency band and determine oscillations.
Therefore, equation (4.21) just represents mid-frequency oscillations of the frequency
response trace. Based on equation (4.21), mid-frequency resonances of transformer winding
frequency response trace are completely dependent on winding inductance as well as series
and shunt capacitances. Any changes in the values of inductance or capacitances as well as
winding length may lead to changes in resonance frequencies.
On the other hand, different types of transformer windings are designed and manufactured
based on voltage level and electromagnetic relations. When a unit-function voltage is
applied to transformer, the initial distribution of voltage is determined entirely by the
capacitive network.
According to the literature ‎[99], the initial impulse voltage distribution intensity is
determined through α parameter. A low α value will lead to a more uniform initial impulse
voltage distribution, whereas a high α value will create severe stress on the upper disks of
the transformer winding. It is worth noting that α parameter can be calculated from
63
equation (4.13) and is equal to equation (4.22) when the capacitive network is considered as
the equivalent circuit of transformer winding:
cg
  (4.22)
cs
Since the initial distribution is controlled entirely by the ratio of capacitances, numerous
studies have been conducted in the past aiming to explore methods to decrease the value of
α in various types of transformer windings ‎[99]. This may be done either by reducing the
shunt capacitance cg or alternatively by increasing the series capacitance cs. In fact, each and
every transformer winding type has its own α value. It is mathematically proven that
interleaved winding shows lower α compared to continuous disk winding ‎[93],‎[99]-‎[100].
According to equation (4.21), significant value for shunt capacitance compared to series
capacitance (large α) will lead to maximum number of resonance frequencies. On the other
hand, large series capacitance with respect to shunt capacitance (small α) results in
minimum number of resonance points. This behaviour was also discussed in ‎[27]. Hence, the
frequency response trace for continuous disk winding has more oscillations in mid
frequencies while interleaved winding has fewer changes over the same frequency band.
This hypothesis will be explored through practical measurements in the next subsections in
order to validate the modelling and calculation carried out in the last Section.
4.4 Verification of Mathematical Calculation Using Practical Measurement
In order to verify mathematical calculation for mid-frequency oscillations of FRA trace, the
glassy model transformer (Fig. 3.9) in the absence of transformer oil was used as the first
test object.
An aluminium foil has been wrapped over the glass casing and grounded to simulate the
metal transformer tank. To carry out the frequency response simulation of manufactured
test object, electrical parameters of the glassy transformer should be calculated and utilized
in mathematical result. This work has been done using the formulas provided in the last
Chapter. Calculated electrical parameters for the test object were validated through practical
measurement using RLC meter and dielectric dissipation factor (DDF) measurement devices.
The maximum applied voltage for DDF measurement was 5 kV to avoid any kind of
undesirable flash-over in the test object.
Afterwards, the frequency response was simulated for the HV winding of the test object
based on mathematical calculation results and also achieved parameters.
64
Note that the significant value for air-core magnetic reluctance of the test object will lead to
small self-inductances for the windings of the glassy transformer. Therefore, the resonance
frequencies in FRA trace for HV and LV windings will be shifted to higher frequencies.
Hence, the upper band limit for FRA simulation was extended from 2 to 20 MHz to monitor
FRA trace oscillations. Practical measurement of FRA trace was performed for the test object
to compare with simulation result. FRA measurements were made by applying a 20 Vpp
swept sinusoidal signal from 20 Hz to 20 MHz. The reference signal was injected through
input-lead of the HV winding and the response signal was recorded through its neutral-lead
(end-to-end measurement). The characteristic impedance of the measurement cables was
50 , the cable braids were grounded through the aluminium tank. The ambient
temperature surrounding the test object was 27 °C. The simulation and measurement
results for the HV winding of the test object are shown in Fig. 4.3.
20
-20
Magnitude [dB]
-40
-60
-80
Simulated
Measured
-100 2 3 4 5 6 7
10 10 10 10 10 10
Frequency [Hz]
Figure 4.3. Simulated and measured frequency response traces for HV winding of the manufactured glassy
transformer.
In the case of FRA trace oscillations, it can be observed that the resonance and anti-
resonance points in the simulated result are ‘synchronized’ with the measured trace.
However, the discrepancy between simulated and measurement results could be due to the
existence of additional loss mechanism, loss in the aluminium tank or maybe it comes
through boundary condition approximation (A1 and A2) in the calculations.
According to Fig. 4.3, moving from 20 Hz to 10 kHz both traces display almost constant value
due to small value of inductive reactance. Moving to higher frequencies, the reactance due to
65
the self-inductance become significant and Vout in equation (2.7) becomes smaller; therefore,
the frequency response magnitude reported in Bode diagram follows a falling trend.
Interaction between self-inductance and capacitances starts in the first anti-resonance point.
Oscillation behaviour for both spectra in the rest of frequency band comes through equation
(4.21).
4.5 Practical Study
4.5.1 Case Study 1
In order to have precise investigation on mathematical calculation and evaluate continuous

and interleaved disk windings frequency responses, experiments were carried out on two
66 kV, 25 MVA transformer windings. The continuous disk winding contains 72 disks with 8
turns per disk while the interleaved winding has 32 disks with 24 turns per disk.
For this study, FRA measurements were made by applying a 20 Vpp sinusoidal signal across
each winding, at various frequencies in the range 20 Hz - 2 MHz. The measurements were
performed for 801 points (frequencies). Input and output impedances of measurement
cables were 50 Ω. The ambient temperature was 29 °C. The test circuit was connected as in
Fig. 4.2(a). Voltage was injected through the winding input end and output voltage was
measured at the winding output end. Tested windings are shown in Figures 4.4(a) and
4.4(b). Note that an aluminium cylinder was fitted inside each winding and grounded to
simulate the earthed core.
Figure 4.5 shows the FRA measurement results for interleaved and continuous disk
windings; the mid-frequency band in frequency response traces has been highlighted by a
dashed-line rectangle. It can be seen that the frequency response of the continuous disk
winding experiences a large number of oscillations within the mid-frequency range
including resonances (maxima) and anti-resonances (minima). These oscillations occur due
to the kr parameter substitution in equation (4.21) where cg >> cs..
66
(a)
(b)
Figure 4.4. Continuous and interleaved disk windings, (a) Continuous winding, (b) Interleaved winding.
-10
-20
-30
Magnitude [dB]
-40
-50
-60
-70
Interleaved Disc Winding
Continuous Disc Winding
-80 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 4.5. Frequency response traces of continuous and interleaved disk windings.
67
In contrast, the interleaved winding displays a monotonously uniform trend in mid

frequencies as it involves significant series capacitance value. On the other hand, in moving
from low-frequencies to mid-frequencies, the first resonance point relies on the interaction
between the winding self-inductance and winding capacitances. Frequency responses of
both continuous and interleaved windings demonstrate a rising trend in high frequencies as
they are more influenced by the shunt capacitance and connections’ resistances in this
frequency range. As the outer metal tank is absent in this particular case, the shunt
capacitance is from only between winding conductors and the earthed aluminium core.
4.5.2 Case Study 2
Practical studies were also carried out on two 45 MVA and 66 MVA transformers to examine
mathematical result for power-rated transformers. Frequency response spectra were
measured for HV windings of specified transformers in Table 4.1. The 45 MVA transformer
has continuous disk winding while the 66 MVA contains interleaved winding. For both
transformers, input signal was injected through phase A and output signal was taken
through the neutral bushing while other transformer terminals were left open circuit
(floating). The measurements were performed for 801 points (frequencies) and FRA traces
were recorded from 2 Hz-2 MHz. Input and output impedances of measurement cables were
50 Ω. The ambient temperature at the time of measurements was 21°C. The FRA traces are
shown in Fig. 4.6.
Table 4.1. Transformer specifications.
Specifications
Capacity Voltage [kV] Vector Group HV winding type
Transformer no.
T1 45 MVA 245/10.5 YnD11 Continuous
T2 66 MVA 245/11.5 YnD11 Interleaved
According to Fig. 4.6, two anti-resonances (minima) are encountered when moving from the
left-hand side to the right-hand side on FRA traces. The first anti-resonance is due to
influence of the transformer middle limb (B) while the second one is caused by the
transformer lateral limb (C). This will be discussed later in Chapter 5. Moving from low
frequencies to mid frequencies shows discrepancy between the two spectra. The continuous
disk winding experiences oscillatory behaviour while the interleaved winding follows a
rising trend in the mid-frequency band as it was discussed earlier.
68
0
Continuous Disc Winding
-10 Interleaved Disc Winding
-20
-30
Magnitude [dB]
-40
-50
-60
-70
-80
-90
-100 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 4.6. FRA traces of HV windings phases A for 45 MVA (continuous disk winding) and 66 MVA (interleaved
disk winding) power transformers.
4.5 Conclusion
Travelling wave and transmission line theories were utilized to analyse the frequency
response trace oscillations in the case of mid frequencies. The mid-frequency resonance
points in FRA trace were verified through mathematical calculations. FRA mid-frequency
oscillation dependency on inductance as well as series and shunt capacitances was explored.
The glassy model transformer was utilized to compare the simulation and measurement
results. Mathematical calculations were validated through practical measurements.
For experiment, a continuous disk winding, an interleaved disk winding as well as two
power transformers were used to find series and shunt capacitance influences on FRA
fluctuations. It was revealed that a winding with lower α will lead to more oscillations in the
mid frequencies of FRA trace while greater α will result in a steady trend in the mid-
frequency band.
To eliminate the impact of the transformer core on FRA spectrum, the modelling was
conducted on an air-core transformer winding and validation was achieved through
measurements on a model winding and also two full-size windings – all with air-core. FRA
measurement on two power transformers also validates the mathematical model,
specifically in the mid-frequency band. It was demonstrated that mid-frequency oscillations
in FRA spectra of power transformer are affected by series and shunt capacitances,
69
significantly. The transformer metal core influence is negligible in terms of mid-frequency

interpretation.
70
Chapter 5. Low Frequency Interpretation of FRA Signature
Chapter 5
Low Frequency Interpretation of FRA

Signature
5.1 Introduction
Mid-frequency oscillations in FRA spectrum was discussed in Chapter 4. It was shown that
the transformer winding structure is the main reason of the resonances and anti-
resonances in this frequency band.
This Chapter aims to interpret the oscillations in the low-frequency band of FRA spectrum.
Based on literatures [23], [102], the low-frequency band oscillations in FRA trace is
affected mainly by the transformer magnetic core. Actually, it refers to the impedance of
the transformer magnetic core. This parameter is affected through the winding
inductance, and the winding inductance itself comes through the transformer core
reluctance.
Therefore, on one hand to interpret the low-frequency band of FRA trace the core
reluctance should be taken into consideration; on the other hand, this parameter is not
simple to measure or calculate as the frequency is swept. Also, different transformers have
different magnetic core laminates, and their characteristics can be quite different.
Specifically when the frequency is swept from 20 Hz to 2 MHz, their behaviour becomes
non-linear. Hence, working on this case requires a technique to derive the frequency
variable magnetic core reluctance. The technique introduced in this Chapter uses reverse
calculation on the FRA signature. Indeed, the information of magnetic core reluctance is
supposed to be a kind of hidden data available in the winding FRA signature.
In addition, each and every winding experiences its own core reluctance. Therefore,
another method is required to practically distinguish the winding magnetic reluctance of a
phase from that of another phase. This method is also discussed in this Chapter.
This Chapter intends to show that the oscillations occurring in the low-frequency band of
71
FRA trace are mainly affected by the middle and lateral core limbs of the other windings.
This means that the first anti-resonance in FRA trace comes through the middle limb
influence, while the second one is affected through the lateral limb. If one of those limbs is
excluded from the FRA setup or blocked, its corresponding anti-resonance will be
eliminated.
At first, this idea is validated by experiment in this Chapter, and then mathematical
approach is used to verify the practical findings. To this end, the frequency response of a
winding in a power transformer is generated mathematically through the reluctance
spectrum of the other limbs and the result is compared with practical measurements. It
reveals that the generated trace is quite similar to the measured spectrum, and its
resonances and anti-resonances are completely matched.
The interaction between the winding shunt capacitance and inductance to produce anti-
resonance in the low-frequency band is another topic for discussion in this Chapter.
In this Chapter, a 10/0.4 kV, 100 kVA three-phase core type transformer is taken as a test
object. Note that to differentiate between the two winding sides; lower-case symbols are
used to denote low voltage parts (a, b, c, n). Similarly, upper-case symbols refer to high
voltage parts (A, B, C).
5.2 Flux Division Theory
5.2.1 Technical Concept
Essentially, in a transformer the alternating current in the primary winding creates a

magnetic flux in the transformer core. The magnetic flux density is given by:
   B(r, t )dA  1N i (5.1)

c w
where ϕ represents the flux, B is the flux density, dAc is an infinitesimal area of the core
cross section, R is the magnetic core reluctance, i denotes the current, and Nw is the total
number of winding turns.
According to Faraday's Law, when a conductor surrounds a time varying magnetic field -
in the core - it can induce a voltage onto the conductor. This induced voltage consequently
will result in an alternating current through the conductor. The induced voltage is given
by:
72
d
e  Nw (5.2)
dt
Presumably there is a load connected to the secondary side of the transformer, meaning
that the induced current will be carried by the load as well.
On the other hand, Lenz’s law is a simple rule to find the direction of induced current.
Lenz’s law implies that the induced current will flow in such a direction so as to oppose
the cause that produces it. Therefore, if a turn-to-turn or disk-to-disk short circuit occurs
in a transformer winding there will be a circulating current in that short-circuited loop
which tends to oppose the flow of magnetic flux in the transformer core – the cause that
produces it. Eventually, the magnetic field initiated by the circulating current in the short-
circuited loop will not allow the transformer main magnetic flux pass through the
corresponding limb. Hence, theoretically the induced voltage across the defected winding
is equal to zero which is approximately verified in practical measurements. Based on this
fact, transformer Flux Division Theory (FDM) has been employed for decades as an on-site
diagnosis test to recognize transformer winding turn-to-turn and disk-to-disk short
circuit.
FDM can be conducted on single-phase or three-phase core type transformers. This
method will yield adequate information about turn-to-turn or disk-to-disk short circuit in
transformer winding on a certain limb.
5.2.2 Flux Division Measurement (FDM)
FDM can be performed on a transformer on HV or LV side. The test voltage in FDM method
must be applied onto phase A, B or C while the other phases are left open circuit. Under
such a circumstance, only one single winding is excited and the main magnetic flux is
generated in the corresponding limb as per Faraday’s law. The generated magnetic flux
will pass through other limbs to complete flux loop.
Magnetic flux passing through other limbs will create an electromotive force on the
windings of other phases and therefore a voltage will be induced. If an excitation voltage is
applied to a HV winding, the algebraic sum of the induced voltages on the other two HV
windings will be equal to the excitation voltage, theoretically. This can be approximately
validated in practical measurements for wye and delta winding connections.
For instance, in transformers with wye connection, if 230 V is injected to HV side phase A,
the sum of induced voltages in HV side winding phases B and C will be around 230 V as
calculated in (5.3):
73
VA (t )  VB (t )  VC (t ). (5.3)
Practical experience shows that discrepancy between the injected voltage (230 V) and
induced voltages can be up to a few volts.
According to Lenz’s law, the induced voltage in other windings would be around zero while
turn-to-turn or disk-to-disk short circuit occurs. It is assumed that short circuit has
happened in a winding that belongs to a phase other than the associated phase of the
excited winding. When a short circuit occurs in a winding then there will not be any flux
flow in the associated limb for that winding and the generated flux will flow through the
other limbs of the core. Hence, there will not be any voltages induced on the windings with
short-circuited disks or turns. This effect is utilized to distinguish the faulty winding from
the others and that is considered to be the main aim of the FDM method.
The magnetic flux flow due to the excited winding of phase A when disk-to-disk short
circuit has occurred in HV side of phase B is shown in Fig. 5.1. For shell type and five-limb
core type transformers, there are some non-wound auxiliary limbs that always carry some
magnetic flux. This is not an obstacle to distinguish the windings with short-circuited
turns or disks. A transformer with five limbs as well as its associated magnetic circuit is
illustrated in Fig. 5.2.
Figure 5.1. Magnetic flux flow due to phase A excitation when short-circuit occurred in phase B.
Magnetic flux division in a five-limb transformer is illustrated in Fig. 5.3 while short circuit
has occurred in HV or LV side of phase b/B. According to Fig. 5.3, measured voltage on the
side phase will be floating around a certain value which depends on the transformer
construction and technical specifications.
In addition, turn-to-turn or disk-to-disk short circuit in transformer winding can change
winding DC resistance values as well as the no-load current. Although altered DC
74
resistance or no-load current cannot be interpreted as internal short circuit in transformer

winding directly. FDM was introduced in this Section in preparation for interpretation of
the FRA low-frequency band using the short circuit technique.
Figure 5.2. Five-limb transformer active part as well as its magnetic circuit.
Figure 5.3. Magnetic flux division in five-limb transformer due to phase a/A excitation and short circuit
occurred phase b/B. Limbs 2, 3 and 4 are assumed as phase A, B and C respectively.
75
5.3 Mathematical and Practical Approach to Interpret Low-frequency

Band
In order to interpret FRA low-frequency oscillations, general interpretation of FRA trace is

discussed first. Then, practical approach and mathematical calculation are utilized to
analyse oscillations.
5.3.1 General Interpretation of FRA Trace
A 10/0.415 kV, 100 kVA three-phase core type transformer was taken as a test object to
examine FRA low frequency oscillations. The HV windings were left open circuit.
Transformer winding frequency response traces of phases a, b and c were measured and
shown in Fig. 5.4 over the frequency range 20 Hz to 2 MHz. The equivalent magnetic
schematic of three-phase transformer is shown in Fig. 5.5. Test object specifications are
provided in Table 5.1.
76
Low Freq. Mid Freq. Hi gh Freq.
« ,
•·•·• LV side phase a
----· LV side phase b
Figure 5.4. Frequency response traces of phases a, b and c.
ot··-----.............
- ., .. - L V side phase c
-.'•'•

-5 • ., ,
.,
.........
jjj' '
:!2. -10
\\
.g \ \ ,,
""~ -15
c \\ I
:::!:
..
-20
\•
\
••
•
!
/ V /
lr! '
\ I
\
•• •
•• •
-25 ••
' I
••..
...............
-30 102 10l 4 5
10 10 10'
Frequency (Hz)
77
Figure 5.5. Equivalent magnetic circuit of three-phase transformer.
Table 5.1. Transformer specifications.
Rated Voltage [V] 10000/415 Impedance Volts (%) 4.17

Rated Power [kVA] 100 Number of Phases 3
Rated Current [A] 5.77/139 Number of Limbs 3
Cooling System ONAN Vector Group Dyn1
According to Fig. 5.4, in the low frequency range, the inductive reactance of the
transformer winding is considerably greater than the capacitive reactance and plays a
pivotal role in trace formation. Hence, the frequency response magnitude as given by (2.7)
follows a falling trend. In fact, the low frequency behaviour of transformer winding
impedance (Zw) is completely dominated by its inductive reactance. Therefore, Zw can be
expressed as 2πfL where f denotes frequency and L is the total winding self-inductance.
Mutual inductances between HV and LV windings do not contribute to this part of
spectrum as the HV winding is left open circuit without carrying any alternating current
during the test on LV winding, However, the turn-to-turn mutual-inductance of LV winding
has significant impact on total self-inductance formation.
Moving from left to right in Fig. 5.4, frequency response traces of lateral windings (a and c)
exhibit two minimal peaks (anti-resonances) whereas the middle winding spectrum
shows one minimal peak (This behaviour is a common pattern of FRA trace for all
transformers having star connections, see Figures 2.12 and 4.6). For frequencies above 10
kHz there is no significant discrepancy between recorded traces. In the case of delta
connection, these two minimal peaks are replaced by a minimal peak, though, it depends
on terminal configurations. In this study, attempt has been made to interpret double anti-
resonance which is quite a common occurrence in all routine transformers.
78
Low frequency anti-resonances can be generally interpreted through interaction of the

winding self-inductance and shunt capacitances.
Precise investigation on magnetic flux flow reveals that the flux generated by the lateral
windings should pass through different limbs. For instance, in terms of FRA measurement
on phase a some part of the flux generated in limb a due to the FRA test setup on phase a
will flow through the transformer middle limb b while the other goes through the outer
limb c. In terms of middle winding FRA test, the generated flux runs through similar
electromagnetic paths (limbs a and c).
Based on this fact, FRA lateral traces exhibit two minimal peaks while the middle
spectrum shows one minimal peak in low frequencies.
In addition, the first minimal peak in lateral traces is associated with the transformer
middle limb while the second one is influenced by the outer limb. In the next subsection,
this hypothesis is validated through practical and mathematical approach.
5.3.2 Practical Approach
According to the flux division theory, if an internal short circuit occurs in a transformer
winding the magnetic field initiated by the circulating current in the short-circuit loop will
not allow the transformer main magnetic flux pass through the corresponding limb (the
limb with the short-circuited winding). Based on this, to prevent the magnetic flux flows
through a specific transformer core limb, a deliberate short circuit could be made between
the input and neutral leads of the transformer winding associated with that limb. This in
turn can model the short circuit loop within the winding, block the flux flow through that
limb and divert the magnetic flux into other limbs. It is thus a convenient technique to
eliminate a predetermined limb from the core magnetic path of winding. It is then possible
to distinguish the transformer core limb influences on the frequency response trace.
To clarify the introduced technique and interpret the low-frequency oscillations in FRA
spectrum, the line and neutral leads of phase C of the test object were short-circuited to
simulate short-circuit loop around the transformer core limb c/C. This created short-
circuit will cause the transformer main magnetic flux unable to flow through the core limb
c/C. Other terminals were left open circuit and the frequency response trace was recorded
for phase a over the frequency band from 20 Hz to 2 MHz. The test setup and the magnetic
core schematic are shown in Fig. 5.6 and Fig. 5.7, respectively. The result in Fig. 5.8 shows
the reference and measured FRA traces for phase a.
79
Figure 5.6. FRA measurement setup for phase a when A and C are short-circuited.
Figure 5.7. Equivalent magnetic circuit for FRA measurement setup of phase a when terminals A and C are
short-circuited (this will block the flux flow in limb c/C).
5
LV side phase a (Short circuit in HV side phase C)
LV side phase a (Reference)
-5
Magnitude [dB]
-10
822.2 Hz
-15
-20
1.101 Hz
825.5 Hz
-25 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 5.8. Reference and measured FRA traces for phase a when limb C is surrounded by short circuit loop.
When scanning across Fig. 5.8 from low frequencies towards high frequencies, it can be
seen that the anti-resonances in the reference trace at 822.2 Hz and 1.101 kHz have been
80
replaced by one anti-resonance at 825.5 Hz in the measured trace. When moving to higher
frequencies, both traces experience similar oscillations. Note that the magnetic flux flows
through transformer core legs a and b when the line and neutral leads of phase C are
short-circuited.
Therefore, the single anti-resonance in the low-frequency band of the measured trace
represents the interaction between self-inductance of winding a due to core magnetic path
of phases a and b, and the total shunt capacitance of phase a. It should be noted that the
discrepancy between 822.2 Hz and 825.5 Hz is negligible and comes through
measurement accuracy.
Afterwards, the line and neutral leads of the winding on phase B were short-circuited to
model the short-circuit within the transformer centre leg. Other terminals were left open
circuit and the frequency response trace for phase a was re-recorded. The test setup and
the magnetic core schematic are shown in Fig. 5.9 and Fig. 5.10, respectively. The result in
Fig. 5.11 shows the reference and measured FRA traces for phase a.
According to Fig. 5.11 moving from left hand side to right hand side on FRA traces the anti-
resonances in the reference trace at 822.2 Hz and 1.101 kHz have been replaced by a
single anti-resonance at 1.101 kHz in the measured trace. Moving to higher frequencies
reveals similar oscillations for both spectra. In this case with phase B short-circuited, the
magnetic flux flows through transformer core legs a and c. Therefore, the single anti-
resonance in the low-frequency band of the measured trace at 1.101 kHz represents the
interaction between the self-inductance of winding a due to the magnetic core path of
phases a and c, and the total shunt capacitance of phase a.
The frequency response traces of phase a due to a deliberate short circuit on B and C are
shown in Fig. 5.12.
Figure 5.9. FRA measurement setup for phase a when B and C are short-circuited.
81
Figure 5.10. Equivalent magnetic circuit for FRA measurement setup of phase a when terminals B and C are
short-circuited (this will block the flux flow in leg b/B).
5
LV side phase a (Short circuit in HV side phase B)
LV side phase a (Reference)
0
1.101 Hz
-5
Magnitude [dB]
-10
-15
822.2 Hz
-20
1.101 Hz
-25 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 5.11. Reference and measured FRA traces for phase a when limb B is surrounded by short circuit loop.
82
5
LV side phase a (Short circuit in HV side phase C)
LV side phase a (Short circuit in HV side phase B)
-5
Magnitude [dB]
-10
-15
-20
1.101 Hz
825.5 Hz
-25 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 5.12. Frequency response traces of phase a (open circuit) and deliberate short circuit on phase B and C.
According to Fig. 5.12, the frequency response trace of phase a due to a deliberate short-
circuit loop in phase C shows a single anti-resonance (in the low frequency region). The
frequency of this anti-resonance coincides with the first anti-resonance of the reference
trace but has a different magnitude. This magnitude discrepancy is initiated by changes in
the equivalent core reluctance.
Similarity, the frequency response trace of phase a due to a deliberate short-circuit loop
on phase B displays another single anti-resonance (in the low frequency region). The
frequency of this anti-resonance corresponds to the second anti-resonance of the
reference trace.
Therefore, it has been practically demonstrated that the centre and lateral legs of the
transformer core have different influence on the FRA spectrum. In the next subsection,
this practical achievement is going to be validated through mathematical calculations.
5.4 Mathematical Approach
Frequency response trace of phase a when the other leads are left open circuit (end-to-end
measurement) is given by (5.4), see Fig. 2.10(a) and equation 2.7.
 Zout 
ka  20log10   (5.4)
 a  Zout 
Z
83
where, ka and Za represent the magnitude of frequency response trace and the transformer
LV winding impedance of phase a, respectively, and Zout is the output impedance depicted
in Fig 2.10. The output impedance is typically 50 Ω in the FRA test setup, thus, if Za is
obtained then the frequency response spectrum can be determined using (5.4). On the
other hand, the frequency response trace of phase a when phase B is short-circuited and
other leads are left open circuit can be calculated using (5.5). Also, (5.5) can be replaced by
(5.6) when phase C is short-circuited.
 
 Zout 
ka  20log10   (5.5)
B( sc)  Za  Zout 
 B( sc) 
 
 
 Zout 
ka  20log10   (5.6)
C( sc)  Za  Zout 
 C( sc) 
 
where, kaB(sc) and kaC(sc) represent the trace magnitude of phase a when phase B and C are
short-circuited respectively and other leads are left open circuit. ZaB(sc) and ZaC(sc) denote
transformer LV winding impedance when phases B and C are short-circuited respectively
and other leads are left open circuit.
The discrepancy between (5.4), (5.5) and (5.6) comes through Za, ZaB(sc) and ZaC(sc). In fact,
various winding impedances will lead to discrepancy in low-frequency band between the
traces. Za, ZaB(sc) and ZaC(sc) are given by:
   
   
 Za   La 
   
   
 Za   2 f  La  (5.7)
 B( sc)   B( sc) 
   
Z   La 
 aC   C 
 ( sc)   ( sc) 
where, La is the transformer winding self-inductance of phase a. LaB(sc) and LaC(sc) denote
transformer winding self-inductances of phase a when phase B and C are short-circuited
respectively and other leads are left open circuit. La, LaB(sc) and LaC(sc) are then given by:
84
   
 L   1 
 a   a 
   
   
 La  N w  a1
2  (5.8)
 B( sc)   B( sc) 
   
L   1 
 aC   aC 
( sc)  
   ( sc) 
Substitution of (5.8) into (5.7) becomes:
   
   1 
 Za   a 
   
  2  1 
 Za   2 fN w aB  (5.9)
 B( sc)   ( sc) 
   
 Z   1 
 aC   aC 
( sc)  
   ( sc) 
Ra and Nw denote equivalent magnetic reluctance and number of winding turns in phase a.
RaB(sc) and RaC(sc) are the equivalent magnetic reluctances of phase a when phase B and C
are short-circuited respectively and other leads are left open circuit (see Fig. 5.13).
According to Fig. 5.5, equivalent magnetic reluctances for (5.9) can be calculated through
(5.10), and (5.11) where Rc ||Rl ≈Rc, and Rl denotes the leakage reluctance.
(a)
85
(b)
(c)
Figure 5.13. Equivalent magnetic circuit of transformer when frequency response trace is measured for phase
a, (a) Normal three-phase, (b) HV winding phase B is short-circuited, (c) HV winding phase C is short-circuited.
(3 2)c2  4c y  22y

a  (5.10)
c   y
 
 aB   
 (sc)  4 2  y 
 2 
2  
 (5.11)
 a   c 
 C
(sc)   
 
Rc and Ry represent the transformer limb and yoke reluctances, respectively. The limb and
yoke reluctances are given as (5.12), independently:
86
 l 
 c 
 c   A
c 
  1   (5.12)
 y  0 c 
   l
y 
 
 A
y 
 
lc and ly denote the mean magnetic path length of the core limb and yoke respectively, Ac is
the limb cross-sectional area, Ay is the yoke cross-sectional area, μ0 is the vacuum
permeability and μc represents the core permeability.
In a reverse procedure, Rc and Ry are calculated based on RaB(sc) and RaC(sc) as (5.13) and
(5.14) and substituted into (5.15) to calculate R’a. R’a is the generated magnetic reluctance
for phase a through deviated (or measured) magnetic reluctances.
2a  a
C( sc) B( sc) (5.13)
c 
2
a a
B( sc) C(sc) (5.14)
y 
2
Then, R’a is given by:
 2 
 aB 
a  a  1 ( sc) 
B( sc) 4  a  (5.15)
C( sc) 
 
 
Therefore, we obtained the magnetic reluctance seen by phase a (R’a) based on RaB(sc) and
RaC(sc). However, RaB(sc) and RaC(sc) are not calculable. Thus, one approach is to use the result
of practical measurement in Fig. 5.12. RaB(sc) and RaC(sc) are given as (5.16) using reverse
calculation on measured traces depicted in Fig. 5.12.
 
 aB   1 
 
fNa2   
 ( sc) 
   2 (5.16)
 a  1
  
 C( sc)  
 
Γ and Λ can be derived through kaB(sc) and kaC(sc) based on measurement results in the last
subsection as given in (5.17) and (5.18).
87
 Zout  Zout
 

 
1  k 

(5.17)
  20  aB
 ( sc) 
10
 Zout  Zout
 


1 k 

(5.18)
  20  aC
 ( sc) 
10
Substitution of (5.16) into (5.4) will lead to the generation of frequency response trace of
phase a as it is given by:
 
1
 2 
 
 fNa2 Zout 
ka  20log10   (5.19)
 1  Zout 
 
 
 a 2 fNa2 
 
The original and generated frequency response traces of phase a are shown in Fig. 5.14.
Figure 5.14. Original and generated frequency response traces of phase a.
According to Fig. 5.14, the frequency response trace of phase a is generated through
deviated traces. The frequency of resonances and anti-resonances in the generated trace
are completely synchronized with the reference spectrum.
88
Different magnetic reluctances for traces in Fig. 5.12 led to different anti-resonance
magnitudes between generated and reference spectra. Hence, the generated trace in Fig.
5.14 shows slightly greater magnitude for anti-resonances compared to reference trace as
it is created through the deviated spectra in Fig. 5.12.
5.5 Effect of Core Configuration on FRA Trace
Practical and mathematical approaches showed that the transformer core reluctance
obtained for each winding can initiate the low-frequency anti-resonance formation in the
FRA trace.
It is also shown that the core reluctance of a specific winding certainly depends on the
magnetic path associated with that winding. Thus, a hypothesis arises that for windings
with similar magnetic core reluctance in primary or secondary side, the frequency
response would be identical specifically in the low-frequency band. This can occur when
the numbers of winding turns are quite equal in different windings.
To examine this hypothesis, the transformer equivalent magnetic circuits in Fig. 5.15 were
considered. Table 5.2 shows the frequency response measurement setup for various
configurations in Fig. 5.15.
Different FRA measurements were conducted based on the setups recommended in Table
5.2. Measured frequency responses show that similar core configurations in test setup for
the transformer under test will lead to identical frequency response traces. In fact, the
frequency response trace of the test setup in Fig. 5.15(a) is completely matched with the
test setup in Fig. 5.15(k) due to identical transformer core reluctances as well as
configurations. Similar frequency response traces are achieved for test setups in Figures
5.15(b) and (h). Also, FRA spectra for Figures 5.15(c), (d), (f) and (g) are completely
matched.
It should be noted that a, b and c have similar number of winding turns. Frequency
response traces for test setup in Figures 5.15(b) and (h) are shown in Fig. 5.17. Figure 5.18
shows frequency response spectra of Figures 5.15(c), (d), (f) and (g) while the traces of
other setups are shown in Fig. 5.4.
89
(a) (b) (c)
(d) (e) (f)
(g) (h) (k)
Figure 5.15. Equivalent magnetic circuits when deliberate short circuit is applied for various limbs.
Table 5.2. FRA setup connections for the configurations in Fig. 5.15.
Phase
a b c n A B C
State
Figure 5.15 (a) Vinput O.C. O.C. Voutput O.C. O.C. O.C.
Figure 5.15 (b) Vinput O.C. O.C. Voutput B A O.C.
Figure 5.15 (c) Vinput O.C. O.C. Voutput O.C. C B
Figure 5.15 (d) O.C. Vinput O.C. Voutput C O.C. A
Figure 5.15 (e) O.C. Vinput O.C. Voutput O.C. O.C. O.C.
Figure 5.15 (f) O.C. Vinput O.C. Voutput O.C. C B
Figure 5.15 (g) O.C. O.C. Vinput Voutput C. O.C. A
Figure 5.15 (h) O.C. O.C. Vinput Voutput B A O.C.
Figure 5.15 (k) O.C. O.C. Vinput Voutput O.C. O.C. O.C.
O.C.=Open Circuit
90
Figure 5.16. Frequency response traces for test setups of Figures 5.15 (b) and (h).
Figure 5.17. Frequency response traces for test setups of Figures 5.15 (c), (d), (f) and (k).
This in turn can be used as a technique for core deformation localization in transformer.
5.6 Shunt Capacitance Influence
Transformer winding capacitance contains series and shunt capacitances. Series

capacitances consist of turn-to-turn and disk-to-disk capacitances whereas shunt
capacitance is the capacitance of winding with respect to the core and to the electrostatic
91
screens/tank. It is well-known that shunt capacitance effects can be explored in the FRA
low frequencies but there is no practical evidence and physical validation on this
hypothesis. This hypothesis is going to be validated in the next subsection.
5.6.1 Practical Approach
To validate shunt capacitance effects on FRA low frequency spectrum, at first the
frequency response of phase b was recorded when input and output leads of phase B were
short-circuited together. Afterwards, this short-circuit was further connected to the
ground and FRA spectrum was measured once more. This experiment should cause
changes to the shunt capacitance of phase b. This in turn results in frequency response
deviation of this phase as well. Figure 5.18 shows FRA measurement results. The reference
FRA spectrum of phase b is also provided for comparison with the measured traces.
Figure 5.18. Frequency response traces for phase b when HV side phase B is left open circuit, short-circuited,
short-circuited and grounded.
According to Fig. 5.18, frequency response of phase b has altered from the reference trace
for the short-circuited and isolated B as well as the short-circuited and grounded B.
For isolated B (Trace (2) in Fig. 5.18), the frequency response trace shows smaller
magnitude for the first minimal peak due to the greater magnetic reluctance as compared
to reference trace. In addition, the minimal peak is slightly shifted to higher frequency due
to the lower self-inductance. For the grounded B (Trace (3) in Fig. 5.18), the first minimal
peak occurred in lower frequency. The entire low-frequency band of FRA spectrum has
also shifted to lower frequencies. Since the self-inductance of winding b has experienced
92
similar values for both isolated and grounded traces (trace (2) and trace (3)), the
discrepancy between the spectra comes from shunt capacitance alteration. This practical
result is going to be validated through physical and mathematical approaches in the next
subsection.
5.6.2 Physical and Mathematical Approach
From a physical point of view, comparing grounded with isolated traces of phase b reveals
that the winding shunt capacitance has changed considerably. This alteration has led the
entire low frequency band of FRA trace to shift to lower frequencies.
On the other hand, the frequency of the first minimal peak in FRA trace is given by:
1
fr low  (5.20)
2 LC
where, fr-low is the anti-resonance frequency, L denotes the total winding self-inductance
and C denotes the winding capacitance. The winding capacitance (C) contains series and
shunt capacitances.
In this circumstance the transformer winding structure was not changed; therefore, the
transformer series capacitance has remained unchanged. Thus, according to Fig. 5.18 and
equation (5.20), the transformer shunt capacitance has increased. This increment impacts
FRA low frequencies, specifically the anti-resonance frequency.
Shunt capacitance topography for the LV winding of phase b is shown in Fig. 5.19. Figure
5.19 (a) demonstrates active part configuration. Figures 5.19(b) and 5.19(c) illustrate the
shunt capacitance configuration for isolated and grounded B.
To validate the measurement result, the shunt capacitance of the phase b is calculated and
compared to isolated and grounded B in (5.21) and (5.22) respectively.
CHLCHT (5.21)
Csh  CLC 
(b) CHL  CHT

Csh  CLC  CHL (5.22)
(b)
where, Csh(b) and Cʹsh(b) are the shunt capacitances of HV winding phase b while the HV side of
phase B is short-circuited and isolated, and grounded respectively (see Fig. 5.19). CLC, CHL and
93
CHT denote the shunt capacitances of LV with respect to the core, the shunt capacitance of HV
with respect to the LV and, the shunt capacitance of HV with respect to the transformer tank.
(a)
LV LV
CHL CHL
CLC CLC
HV HV
Core Core
Tank Tank
LV LV
CLC CHL CLC CHL

Csh(b) HV C’sh(b)
CHT
(b) (c)
Figure 5.19. Active part and related shunt capacitances, (a) Active part upper view schematic, (b) Shunt
capacitance configuration for b where HV side phase B is just short-circuited and isolated from the ground, (c)
Shunt capacitance configuration for b where HV side phase B is grounded.
From a physical point of view:
CHL , CLC and CHT  0 (5.23)
Therefore:
94
CHLCHT (5.24)
Csh  CLC  
 Csh  CLC  CHL
(b) CHL  CHT (b)
Validating (5.24) is based on the following expressions:
CHLCHT (5.25)
CLC   CLC  CHL
CHL  CHT
CLCCHL  CLCCHT  CHLCHT  CHLCLC  C 2  CHT CLC  CHLCHT

HL
(5.26)
(5.27)
0  C2
HL
Equation (5.27) is always true. Therefore, the assumption in (5.24) is true and:

Csh  Csh (5.28)
(b) (b)
According to (5.28), the shunt capacitance of grounded connection is greater than that of the
isolated case, mathematically. Therefore, based on (5.20) the frequency response trace of
phase b experienced an anti-resonance in lower frequency as compared to the isolated one. In
addition, it is clarified that the first minimal peak can be interacted through self-inductance
and shunt capacitance. Therefore, the shunt capacitance influences the low frequency band of
the FRA trace.
5.7 Conclusion
Interpretation of the low frequency band of FRA trace was discussed. Mathematical calculation
and practical measurements showed that the first minimal peak in the FRA spectrum comes
through the transformer middle limb while the second one is influenced by the lateral limb. To
validate the mathematical and practical approach, one of the winding frequency response
trace for the test object was generated through deviated traces.
According to the data obtained in this Chapter, similar equivalent magnetic circuits for FRA
setup will lead to identical FRA traces. This concern was investigated through practical
measurement. This in turn suggests a new prognosis technique for transformer core defect
localization. The reason lies in the fact that any deformation or defect in the transformer core
95
will cause the entire transformer FRA trace to change, accordingly. This should be taken into
account in FRA result interpretation.
In addition, the shunt capacitance influence was studied and discussed through mathematical
and practical approach. It was found that the shunt capacitance effects can change the low
frequency minimal peak position. It also has considerable effect on the frequency response
trend and its approach in low frequencies.
In summary, it is worth noting that the transformer winding inductance is not expected to
increase. The winding inductance is strictly related to the transformer magnetic reluctance.
Transformer design normally aims to achieve the optimum (low value) magnetic reluctance.
Thus each and every deformation or undesirable lamination in the transformer core sheets
will cause the magnetic reluctance to increase. Accordingly, the winding inductance will
experience lower value.
Reduction in winding inductance will in turn cause the first anti-resonance in the low-
frequency band to move to higher frequencies. Therefore, the shifting toward the right hand
side in the low frequency band can be interpreted through inductance reduction or shunt
capacitance increment. If this deviation is limited to the low-frequency band until the first anti-
resonance, the winding inductance alteration can be concluded. Otherwise, deviation in FRA
spectrum from the first minimal peak towards higher frequencies indicates changes in the
shunt capacitance.
96
Chapter 6. Axial and Radial Deformation of Transformer Winding
Chapter 6
Axial and Radial Deformation of

Transformer Winding
6.1 Introduction
Mechanical defects in transformer winding such as axial or radial deformation may be due
to short circuit currents, earthquakes, careless transportation between sites, explosion of
combustible gases accumulating in the transformer oil, etc. They can cause changes to the
mutual-inductance, series and shunt capacitances and eventually FRA trace deviation. This
Chapter is specifically concerned with symmetrical and asymmetrical axial deformation of
a disk in transformer winding. An analytical approach is provided to calculate self- and
mutual-inductance variation under such circumstances. Also, the radial deformation in
transformer winding is modelled through the “buckling” and discussed analytically.
Afterwards, a numerical example is provided on a model winding.
In the case of capacitance calculation in radial deformation, the 2D Finite Element Method
(FEM) is also used to evaluate capacitance variation and the results are compared to the
analytical approach.
To explore the influence of winding parameters variation on FRA spectrum deviation, the
inductance, series and shunt capacitances, resistance, conductance to ground and turn-to-
turn conductance in transformer winding are changed and the frequency response is
simulated for the glassy model transformer presented in Chapter 4. The reason of FRA
spectrum deviation in each case is then discussed physically. To verify the simulation
result for each and every case, practical measurements are performed and deviated parts
in the measured FRA trace are compared to simulation results. The practical experiments
conducted verify the simulation result as well as the modelling.
97
6.2 Axial Deformation and Its Impacts on Winding Parameters
It is possible for axial deformation to occur symmetrically or asymmetrically for one or

several disks of a transformer winding.
Symmetrical axial deformation means that when one side of a disk has moved up closer to
the upper disk, the other side is moved down closer to the lower disk as illustrated in
Fig.6.1. In this condition, the axes of the deformed disk will be inclined to other disks.
Asymmetrical axial deformation of a disk refers to the situation when one side of the disk
moves closer to the upper disk whereas the outermost conductor on the opposite side
remains unchanged in original state as depicted in Fig. 6.2. Figures 6.1(a) and 6.2(a) show
the deformation patterns, and Figures 6.1(b) and 6.2(b) illustrate the implementation of
the windings.
Indeed the distance between the transformer winding disks (inter-disk distance) can
change due to the axial deformation. This causes the turn-to-turn mutual-inductance to
change accordingly, and further to that the total self-inductance is altered. An analytical
approach to investigate the mutual-inductance variation under axial deformation is
discussed hereinafter.
6.2.1 Mutual Inductance of Circular Filaments Whose Axes Are Inclined to One
Another
6.2.1.1 Symmetrical Axial Deformation of a Disk
The mutual-inductance of circular filaments whose axes are inclined to one another is
given by ‎[85]:
M   R0M cos  R0 f RbRa cos H (6.1)

m
where M is calculated as (3.21), θ is illustrated in Fig.6.1, and R0 is a function of θ and given

by:
P ( ) 1 4 P ( ) 5 6 P ( )
1  1  2 P3( ) 3   P5( ) 5   P7 ( ) 7  ...
R0  4  8  64 
1  1  2 P3( )  1  4 P5( )  5  6 P7 ( )  ...
4 8 64 (6.2)
where:  2   2 ,  2   2 ,   sin 1  X  ,   cos

2 2  
1  1  R
 a
98
Note that X denotes the axial displacement of the outermost turn. The formulas for Pm(μ)
and P'm(ν) are provided through (B.1) and (B.2) in Appendix B ‎[78]. In the case of
transformer winding, the parameter ψ generally takes a value between 0.6 and 0.9. Hence
for convenience, in the range of 0.6 < ψ < 0.9 some of pre-calculated values of R0 are
provided in Table B.2 ‎[85] in Appendix B.
Based on Table B.2, R0 is a value less than 1 for routine transformer windings. In addition,
cosθ ≤ 1; hence, M' ≤ M (M was discussed in Chapter 3). Therefore, the mutual inductances
between the winding turns for the case of symmetrical deformation of a disk take smaller
values as compared to the normal winding. This in turn influences the total self-inductance
of the winding and makes it smaller.
99
Ra
θ
Rb
(a)
Air-core
W
1 2 3 4 4 3 2 1
8
θ 7
6
5
5
6
7
8
d
9 10 11 12 12 11 10 9
16 15 14 13 13 14 15 16
(b)
Figure 6.1. Symmetrical axial deformation of a disk, (a) Axial deformation pattern, (b) Deformed disk.
100
Ra
θ
D
d
Rb
(a)
Air-core
W
1 2 3 4 4 3 2 1
D'
8 7
θ' 6 5
5 6
d'
d'
7 8
D
d
9 10 11 12 12 11 10 9
16 15 14 13 13 14 15 16
(b)
Figure 6.2. Asymmetrical axial deformation of a disk, (a) Axial deformation pattern, (b) Deformed disk.
101
6.2.1.2 Asymmetrical Axial Deformation of a Disk
The mutual-inductance between the turns for asymmetrical axial deformation of a disk is
given by:
M   R0M a cos   R0 f  RbRa cos  H (6.3)

m
According to Fig. 6.2, the distance between the second and third disks has been changed
from d to D, while D > d. Hence, χ in (6.2) should be replaced by χdown for mutual inductance
calculation between the turns in the second and third disks, and given as follows:

down  D  d  Ra sin (6.4)
Rb Rb
For the mutual-inductance between disk two and disk one, χup is then defined as (6.5) and
represents the ratio of the distance between the first and second disks and Rb:
 
up  D  d  Ra sin (6.5)
Rb Rb
Other parameters in (6.2) should be also replaced by:
 2   2 ,  2   x 2 ,    sin 1  X  ,    cos
2 2  
(6.6)
1  x 1  x  2Ra 
χx can take χdown or χup in (6.6) for the mutual inductance calculation of the second disk with
respect to the third or first disk, respectively. In addition, μ can be replaced by μ', κ2 by κ'2
and ν2 by ν'2 in (6.6) and k by kdown or kup in (3.22) and (6.3), accordingly.
Since χup < χ < χdown , therefore, kup < k < kdown and fm(up) > f > fm(down). Hence, we define A1=
fm(up) – fm > 0, and A2 = fm(down) – fm < 0. Based on (6.3), if fm(down) < fm(up) then |A1| >|A2| and
eventually the total value of fm'' for the asymmetrical deformation would be less than the
mutual inductance coefficient, fm, in the normal winding. This factor causes M'' to
experience larger value than M (see equation (3.23) to recall fm calculation).
On the other hand, as discussed earlier R0 is a value less than 1 for typical transformer
winding. In addition, cosθ' ≤ 1. Hence, M'' can take a value less than M.
Therefore, two factors (cosθ and R0) will reduce M'' whereas one factor (fm'') would
increase it. Thus, it is difficult to argue analytically on the magnitude of M'' as compared to
M. In order to determine which factors are more effective in changing their relative
magnitude, a numerical example is provided in Section 6.4 and the results are discussed.
102
6.2.2 Capacitances of Circular Filaments Whose Axes Are Inclined to One Another
6.2.2.1 Series Capacitance
Series capacitance comprises total turn-to-turn (Ctt) and total inter-disk capacitances (Cd).
Turn-to-turn transformer winding capacitance can be influenced slightly by the axial
symmetrical and asymmetrical deformation of a winding’s disk.
For the deformation illustrated in Fig. 6.1, equation (3.26) for the turn-to-turn capacitance
of the second disk is obtained as:
 h  w tan   2 t (1  tan  ) 
Ctt   t  0 D   . (6.7)
 2t 
Ctt Ctt Ctt 1 l

1 2 3 4
U 2 2U
2N 3
4
U
U Cdd 5 2N
8
7 6
tt 6 7
5
tt 8
tt U
2N 9
10
U
9 10 11 12 11
12
Ctt Ctt Ctt U
13 2N
U 14
15
16 15 14 13 16
0
Figure 6.3. Cross-section overview and voltage distribution along deformed winding.
The voltage distribution in along transformer winding length (voltage per turn) is mostly
influenced by number of winding turns as well as winding height. Therefore in the case of
inter-disk capacitance, if assuming the voltage distribution along the pair of disks still
remained uniform, in axial deformation the total inter-disk capacitance (Cd) will not
change (see Fig. 6.3).
Base on this, the total series capacitance in Fig. 6.1 will slightly reduce. This reduction
could be negligible as the transformer winding disks are practically close together and
tanθ experiences small value. Similar calculations can be performed for Fig. 6.2.
103
6.2.2.2 Shunt Capacitance
The shunt capacitance between the winding and the container (metal tank) can experience
similar value as given by equation (3.34). This is because the outward structural
configuration of winding is still similar to the normal winding.
6.3 Radial Deformation and Its Impacts on Winding Parameters
In this study, radial deformation of transformer winding is modelled through free-buckling

as shown in Fig. 6.4, where the winding exterior has faced buckling towards metal tank.
Inward buckling occurs when the circumference of the winding disk is bent towards the
disk centre. Based on this, the winding inductance as well as series and shunt capacitances
are going to be discussed in this Section.
180
120
r2
240 r1
r0
60
300
180
120
240
Winding
60
300
Metal container
0
Figure 6.4. Radial deformation schematic (free-buckling).
6.3.1 Self and Mutual Inductances in Radial Deformation
Analytical calculation of the self and mutual inductances for the transformer winding
which has experienced inward or outward buckling is quite a challenging task to address.
Most of the literatures have referred to the work and equations presented by Maxwell in
104
‎[78]. Work by Chattopadhyay in ‎[103] on super conductors has tried to address this issue
through analytical approach.
In this thesis, it has been tried to improve this issue analytically to move one step forward
and find the loop inductance value under deformation. To this end, the provided equations
by Maxwell ‎[78] and Paul ‎[104] for the calculation of self and mutual inductances of the
circular loops are used and extended to determine the self and mutual inductances of the
non-circular filaments. However, the resultant integrals obtained in some cases are quite
non-linear and should be solved numerically. Thus, numerical calculation is conducted to
complete the inductance values calculations in radial deformation. In detail, this task is
provided separately in Appendix B due to significant volume of calculations.
Furthermore, radial displacement is a different concept as compared to radial

deformation, but in some literatures ‎[102] it has been considered and discussed as radial
deformation (see Fig. 6.5). In this case, the mutual inductance of circular elements with
parallel axes is given by ‎[85]:
M   M p F (6.8)
rd R d
θ"
θ" rd d
R ρ
Figure 6.5. Radial displacement pattern of a disk.
where, Mp is calculated as M with a distance rd between the disks and F is given as Table
B.3 in Appendix B.
Since rd > d, therefore Mp < M. According to Table B.3, F can take a value larger or smaller
than 1. Based on this and also data in Table B.3, M''' will not change significantly for the
usual buckling of transformer winding.
105
6.3.2 Series and Shunt Capacitances in Radial Deformation
6.3.2.1 Series Capacitance
According to equations (3.26) through (3.30), it can be concluded that the series
capacitance including turn-to-turn and inter-disk capacitances of the buckled winding
shown in Fig. 6.4 is not changed significantly because the dimensions still remained
unchanged for these parameters.
6.3.2.2 Shunt Capacitance
Shunt capacitance calculation for the deformed winding is discussed in ‎[13]; however, the
provided formula requires to be modified. Therefore, a more accurate approach is
suggested here.
The electric field generated by the transformer winding is not uniform across the
deformed turns. Therefore, calculation of the shunt capacitance requires the electric field
determined by finite element method. This approach is demonstrated in the next
subsection through a numerical example, and the result is then compared with the
analytical approach.
In the case of analytical approach, if assuming a uniform electric field across the deformed
section (highlighted in Fig 6.4), the equivalent shunt capacitance can be calculated through
the summation of the shunt capacitance of the normal section, Cnorm, as given by (6.9)
paralleled with the shunt capacitance of the deformed section, Cdeform, as given by (6.10):
2  0 r H w
Cnormal   2 d
(2 

) r  (6.9)
ln  2 
 r1 
2
 0 r H w
Cnormal    d
0  r2  (6.10)
ln  
 r1  0.5r (cos( )  1) 
where, η is the ratio of entire trigonometric circular span (2π) over the deformation span
(rad) as illustrated in Fig. 6.4, Hw is the winding height, and r' represents the deformation
radius. The total shunt capacitance of the winding that had buckling is then obtained as:
Cg  Cnormal  Cdeform (6.11)
106
6.4 Numerical Example
In this Section, a numerical example is provided to examine the inductance and

capacitance variation in winding deformation.
It is assumed that the model winding illustrated in Fig. 3.1 has single strand conductor.
Radial dimension of conductor is w = 7 mm, axial dimension is h = 11 mm, inter-disk
distance is δd = 6 mm, thickness of paper insulation is δt = 0.5 mm, relative permittivity of
paper insulation is εt = 3.2, vacuum permittivity is ε0, mean radius of winding disk is R=
280 mm, mean radius of the tank is R= 400 mm, and winding as well as tank height are 62
mm and 100 mm, respectively. The winding has four disks and four conductors per disk.
Note that centre to centre distance for pair of disks is d = 5.5 + 5.5 + 6 = 17 mm.
6.4.1 Axial Deformation of a Disk
Axial deformation of model winding is assumed to occur for the second disk such as to
what happened in Figures 6.1 and 6.2. Hence, the maximum axial displacement in Fig. 6.1
and Fig. 6.2 for the conductor numbered 8 would be X = 6 mm. Therefore, the inductance
and capacitance of the model are calculated as follow using the mentioned formulas.
6.4.1.1 Inductance Calculation
Equations (3.19) and (3.21) were used to calculate the self and mutual inductances. Thus,
the detailed inductance matrix (μH) of the normal winding obtained is:
1.8221 1.3679 1.0997 0.9387 0.8500 0.9385 1.0190 1.0653 0.8128 0.7910 0.7564 0.7131 0.6006 0.6277 0.6498 0.6655 
 
1.3679 1.7354 1.3250 1.0630 0.9066 0.9863 1.0310 1.019 0.791 0.7845 0.7631 0.7291 0.6036 0.6253 0.6409 0.6498 
1.0997 1.3250 1.6507 1.284 0.9529 0.9969 0.9863 0.9385 0.7564 0.7631 0.7565 0.7354 0.6010 0.6164 0.6253 0.6277 
 
0.9387 1.0630 1.2840 1.5682 0.9629 0.9529 0.9066 0.8500 0.7131 0.7291 0.7354 0.7287 0.5922 0.6010 0.6036 0.6006 
0.8500 0.9066 0.9529 0.9629 1.8221 1.3679 1.0997 0.9387 0.8500 0.9066 0.9529 0.9629 0.8128 0.7910 0.7564 0.7131 
 
0.9385 0.9863 0.9969 0.9529 1.3679 1.7354 1.3250 1.0630 0.9385 0.9863 0.9969 0.9529 0.7910 0.7845 0.7631 0.7291 
1.019 1.0310 0.9863 0.9066 1.0997 1.3250 1.6507 1.2840 1.0190 1.0310 0.9863 0.9066 0.7564 0.7631 0.7565 0.7354 

1.0653 1.019 0.9385 0.8500 0.9387 1.0630 1.2840 1.5682 1.0653 1.0190 0.9385 0.8500 0.7131 0.7291 0.7354 0.7287 
Leq 0.8128 
0.791 0.7564 0.7131 0.8500 0.9385 1.019 1.0653 1.8221 1.3679 1.0997 0.9387 0.8500 0.9385 1.019 1.0653  (6.12)
0.791 0.7845 0.7631 0.7291 0.9066 0.9863 1.0310 1.0190 1.3679 1.7354 1.3250 1.0630 0.9066 0.9863 1.0310 1.019 
 
0.7564 0.7631 0.7565 0.7354 0.9529 0.9969 0.9863 0.9385 1.0997 1.3250 1.6507 1.284 0.9529 0.9969 0.9863 0.9385 
0.7131 0.7291 0.7354 0.7287 0.9629 0.9529 0.9066 0.8500 0.9387 1.0630 1.284 1.5682 0.9629 0.9529 0.9066 0.8500 
 
0.6006 0.6036 0.6010 0.5922 0.8128 0.791 0.7564 0.7131 0.8500 0.9066 0.9529 0.9629 1.8221 1.3679 1.0997 0.9387 
 
0.6277 0.6253 0.6164 0.6010 0.791 0.7845 0.7631 0.7291 0.9385 0.9863 0.9969 0.9529 1.3679 1.7354 1.3250 1.0630 
0.6498 0.6409 0.6253 0.6036 0.7564 0.7631 0.7565 0.7354 1.019 1.0310 0.9863 0.9066 1.0997 1.3250 1.6507 1.2840 
 
0.6655 0.6498 0.6277 0.6006 0.7131 0.7291 0.7354 0.7287 1.0653 1.0190 0.9385 0.8500 0.9387 1.0630 1.284 1.5682 
For the symmetrical axial deformation occurred in the second disk (Fig. 6.1), the
inductance matrix is calculated as (6.13) using equation (6.1), while this matrix for the
asymmetrical deformation (Fig. 6.2) is obtained as (6.14) using equation (6.3):
107
1.8221 1.3679 1.0997 0.9387 0.8483 0.9365 1.0168 1.0629 0.8128 0.791 0.7564 0.7131 0.6006 0.6277 0.6498 0.6655 
 
1.3679 1.7354 1.3250 1.0630 0.9047 0.9842 1.0287 1.0168 0.791 0.7845 0.7631 0.7291 0.6036 0.6253 0.6409 0.6498 
1.0997 1.3250 1.6507 1.284 0.9508 0.9947 0.9842 0.9365 0.7564 0.7631 0.7565 0.7354 0.6010 0.6164 0.6253 0.6277 
 
0.9387 1.0630 1.284 1.5682 0.9607 0.9508 0.9047 0.8483 0.7131 0.7291 0.7354 0.7287 0.5922 0.6010 0.6036 0.6006 
0.8483 0.9047 0.9508 0.9607 1.8221 1.3679 1.0997 0.9387 0.8483 0.9047 0.9508 0.9607 0.8110 0.7893 0.7548 0.7117 
 
0.9365 0.9842 0.9947 0.9508 1.3679 1.7354 1.3250 1.0630 0.9365 0.9842 0.9947 0.9508 0.7893 0.7827 0.7614 0.7276 
1.0168 1.0287 0.9842 0.9047 1.0997 1.3250 1.6507 1.284 1.0168 1.0287 0.9842 0.9047 0.7548 0.7614 0.7548 0.7338 

1.0629 0.7271 
L
eq 0.8128
1.0168 0.9365 0.8483 0.9387 1.0630 1.284 1.5682 1.0629 1.0168 0.9365 0.8483 0.7117 0.7276 0.7339

0.791 0.7564 0.7131 0.8483 0.9365 1.0168 1.0629 1.8221 1.3679 1.0997 0.9387 0.8500 0.9385 1.019 1.0653  (6.13)
0.791 0.7845 0.7631 0.7291 0.9047 0.9842 1.0287 1.0168 1.3679 1.7354 1.3250 1.0630 0.9066 0.9863 1.0310 1.0190 
 
0.7564 0.7631 0.7565 0.7354 0.9508 0.9947 0.9842 0.9365 1.0997 1.3250 1.6507 1.284 0.9529 0.9969 0.9863 0.9385 
0.7131 0.7291 0.7354 0.7287 0.9607 0.9508 0.9047 0.8483 0.9387 1.0630 1.284 1.5682 0.9629 0.9529 0.9066 0.8500 
 
0.6006 0.6036 0.6010 0.5922 0.8110 0.7893 0.7548 0.7117 0.8500 0.9066 0.9529 0.9629 1.8221 1.3679 1.0997 0.9387 
 
0.6277 0.6253 0.6164 0.6010 0.7893 0.7827 0.7614 0.7276 0.9385 0.9863 0.9969 0.9529 1.3679 1.7354 1.3250 1.0630 
0.6498 0.6409 0.6253 0.6036 0.7548 0.7614 0.7548 0.7338 1.019 1.0310 0.9863 0.9066 1.0997 1.3250 1.6507 1.2840 
 
0.6655 0.6498 0.6277 0.6006 0.7117 0.7276 0.7338 0.7271 1.0653 1.019 0.9385 0.8500 0.9387 1.0630 1.2840 1.5682 
1.8221 1.3679 1.0997 0.9387 0.8495 0.9380 1.0184 1.0647 0.8128 0.7910 0.7564 0.7131 0.6006 0.6277 0.6498 0.6655 
 
1.3679 1.7354 1.3250 1.0630 0.9061 0.9857 1.0304 1.0184 0.791 0.7845 0.7631 0.7291 0.6036 0.6253 0.6409 0.6498 
1.0997 1.3250 1.6507 1.2840 0.9523 0.9963 0.9857 0.9380 0.7564 0.7631 0.7565 0.7354 0.6010 0.6164 0.6253 0.6277 
 
0.9387 1.0630 1.284 1.5682 0.9623 0.9523 0.9061 0.8495 0.7131 0.7291 0.7354 0.7287 0.5922 0.6010 0.6036 0.6006 
0.8495 0.9061 0.9523 0.9623 1.8221 1.3679 1.0997 0.9387 0.8495 0.9061 0.9523 0.9623 0.8123 0.7905 0.7560 0.7127 
 
0.9380 0.9857 0.9963 0.9523 1.3679 1.7354 1.3250 1.0630 0.9380 0.9857 0.9963 0.9523 0.7905 0.7840 0.7626 0.7287 
1.0184 1.0304 0.9857 0.9061 1.0997 1.3250 1.6507 1.284 1.0184 1.0304 0.9857 0.9061 0.7560 0.7627 0.7560 0.7350 

1.0647 0.7283 
L
eq 0.8128
1.0184
0.791
0.9380
0.7564
0.8495
0.7131
0.9387
0.8495
1.0630
0.9380
1.284
1.0184
1.5682
1.0647
1.0647
1.8221
1.0184
1.3679
0.9380
1.0997
0.8495
0.9387
0.7127
0.8500
0.7287
0.9385
0.7350
1.019 1.0653 
 (6.14)
0.791 0.7845 0.7631 0.7291 0.9061 0.9857 1.0304 1.0184 1.3679 1.7354 1.3250 1.0630 0.9066 0.9863 1.0310 1.0190 
 
0.7564 0.7631 0.7565 0.7354 0.9523 0.9963 0.9857 0.9380 1.0997 1.3250 1.6507 1.284 0.9529 0.9969 0.9863 0.9385 
0.7131 0.7291 0.7354 0.7287 0.9623 0.9523 0.9061 0.8495 0.9387 1.0630 1.284 1.5682 0.9629 0.9529 0.9066 0.8500 
 
0.6006 0.6036 0.6010 0.5922 0.8123 0.7905 0.7560 0.7127 0.8500 0.9066 0.9529 0.9629 1.8221 1.3679 1.0997 0.9387 
 
0.6277 0.6253 0.6164 0.6010 0.7905 0.7840 0.7626 0.7287 0.9385 0.9863 0.9969 0.9529 1.3679 1.7354 1.3250 1.0630 
0.6498 0.6409 0.6253 0.6036 0.7560 0.7626 0.7560 0.7350 1.019 1.0310 0.9863 0.9066 1.0997 1.3250 1.6507 1.2840 
 
0.6655 0.6498 0.6277 0.6006 0.7127 0.7287 0.7350 0.7283 1.0653 1.0190 0.9385 0.8500 0.9387 1.0630 1.2840 1.5682 
As compared to (6.12), those elements which have changed due to the winding axial
deformation are highlighted by enclosing rectangles in (6.13) and (6.14).
The data of (6.15), (6.16) and (6.17) are obtained as the overall inductance matrix through
the summation of elements in (6.12), (6.13) and (6.14), respectively. These results
represent the inductance matrix in (3.1).
 20.9330 15.3627 12.0587 9.9310 

 
15.3627 20.9330 15.3627 12.0587 
Leq   (6.15)
12.0587 15.3627 20.9330 15.3627 
 
9.9310 12.0587 15.3627 20.9330 
 20.9330 15.3296 12.0587 9.9310 

 
15.3296 20.9330 15.3296 12.0329 
 
Leq (6.16)
12.0587 15.3296 20.9330 15.3627 
 
9.9310 12.0328 15.3627 20.9330 
108
 20.9330 15.3545 12.0587 9.9310 

 
15.3545 20.9330 15.3546 12.0517 
  
Leq (6.17)
12.0587 15.3546 20.9330 15.3627 
 
9.9310 12.0517 15.3627 20.9330 
According to (6.13) and (6.14), the elements in Mab, Mba, Mbc, Mcb, Mbd and Mdb are changed
due to the axial deformation of the second disk. In fact, all related mutual inductances to
the second disk have changed, while other elements have remained unchanged.
Furthermore, the self-inductance of the second disk (LB) has not changed, while the total
self-inductance of the model winding (Leq) has altered.
In addition, the variation of mutual inductances as well as total self-inductance for the
symmetrical axial deformation is more than asymmetrical deformation. This in turn means
that the deformation angle, θ, is quite a significant parameter in inductance variation.
6.4.1.2 Capacitance Calculation
The series capacitance of the normal winding of (Fig. 3.1) is:
Cs  67.950 pF (6.18)
This value was obtained as (6.19) and (6.20) for the configurations in Fig. 6.1 and Fig. 6.2,
respectively.
Cs  67.827 pF (6.19)
Cs  67.889 pF (6.20)
The shunt capacitance will remain constant as the outward configuration has not changed
much.
6.4.1.3 Practical Experiment on Axial Deformation
The inductance reduction was discussed in the last subsections for the winding having an
axially deformed disk. To examine this result, a 66 kV, 25 MVA air-core interleaved
winding was used to conduct an experiment on axial deformation. This winding contains
32 disks with 24 turns per disk (examined test object in Chapter 4). At first, frequency
response end-to-end open circuit measurement was performed on the original
(undeformed) winding. The FRA spectrum was recorded over the frequency range of 20
Hz - 2 MHz. Afterwards, the fourth disk of the winding was deformed axially with its
109
outermost turn shifted toward the upper disk, as shown in Fig. 6.6. This was achieved by
inserting a plastic wedge to simulate the asymmetrical axial deformation. The winding
frequency response was re-recorded and the two spectra (before and after deformation)
are shown in Fig. 6.7.
Figure 6.6. Axially deformed interleaved winding.
0 -25
-30
-35
-10
Magnitude [dB]
-35
-40 -40
-20
Magnitude [dB]
-45
5
Magnitude [dB]
-45 Frequency [Hz]

10
-30
-50
-55 -40
Frequency Bankd: 20 Hz - 2 MHz
After Axial Deformation
Fingerprint
-60
10-50
4.2 4.3 4.4 4.5
Frequency 10
Bankd: 20 Hz[Hz]
Frequency
10
- 2 MHz 10
After Axial Deformation

Fingerprint
-60 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 6.7. Reference and measured frequency response spectra for deformed interleaved windings.
As achieved in Chapter 5, it was shown that within the FRA spectrum of a transformer
winding, the first minimal peak (anti-resonance) in the low frequencies is initiated
through the interaction between the winding inductive and capacitive reactances. Thus,
study on this anti-resonance will help to estimate which winding’s parameter has altered.
Study on other frequency bands of the FRA spectrum can also narrow down the
investigations, help to analyse the first anti-resonance better, and improve the diagnosis.
According to Fig. 6.7, moving from lower towards higher frequencies, it can be seen that
the spectra in the very low-frequency region are almost matched. After around 1 kHz, the
discrepancy becomes noticeable in the low-frequency band while the first minimal peak
110
shows FRA spectrum deviation due to the axial deformation. This in turn could be due to
the inductance or shunt capacitance reduction of the winding. In addition, mid-frequency
oscillations have altered and resonance and anti-resonances are randomly shifted. Since
the mid-frequency oscillations are initiated through the mutual-inductance within the
total winding inductance as well as series and shunt capacitances, this alteration tends to
support a hypothesis that the mutual-inductance or total capacitance may have been
changed. Furthermore, close matching of the FRA spectra in the high and very high
frequency regions indicates that deviation in spectra comes through inductance rather
than capacitance variation. Available discrepancy in FRA spectra in low frequencies can
also validate this hypothesis. In fact, the low-frequency band of the FRA spectrum is
mainly affected by winding inductance and deviation in this region comes through self- or
even mutual-inductance alteration. This is the fact that the first anti-resonance is shifted to
higher frequencies due to the inductance reduction in this winding, see equation (5.20).
Practical measurement results of this study on a single air-core interleaved winding verify
the analytical approach and numerical calculation on axial deformation.
6.4.2 Radial Deformation along the Winding
Here, it is assumed that radial deformation caused part of the model winding to be
stretched towards the metal tank as shown in Fig. 6.4. The mean radius of deformation is
r'= ±50 mm, and the deformation angle is φ= π/2.
6.4.2.1 Inductance Calculation
Self- and mutual-inductance variation due to the radial deformation in transformer

winding is discussed in details in Appendix B.
6.4.2.2 Capacitance Calculation
In this particular case, the shunt capacitance of the winding can change significantly due to
the radial deformation, while the series capacitance will remain almost constant.
The shunt capacitance is simulated and calculated for a buckled winding similar to Fig. 6.4
using finite element method as well as the formula presented in this Chapter. The
parameters of the simulated winding are same as those presented in the numerical
example shown in Fig. 6.8. Simulated and analytical results are presented in Table 6.1. The
data obtained in Table 6.1 reveal that the analytical approach to calculate the shunt
capacitance is reasonably accurate. However, this accuracy might not be adequate for
complicated radial deformations. For both finite element and analytical approaches, the
111
winding which is bent inward experiences less shunt capacitance than that winding bent
outward.
Figure 6.8. The modelled winding through finite element (r1= 280 mm, r2 = 400 mm, r'= ±50 mm, φ= π/2).
Table 6.1. Calculated capacitance between the winding and metal container (tank).
Condition Finite Element Analytical Calculation

Normal Winding 9.6893 pF 9.6658 pF
Bent outward 10.6976 pF 10.5604 pF
Bent inward 9.2218 pF 9.1997 pF
6.5 A Summary on Axial and Radial Deformations
Modelling of transformer windings due to radial and axial deformations is considered a

significant challenge for those researchers studying transformer winding through the
detailed model. This Chapter is focused on inductance and capacitance variations due to
winding deformation. A winding model was proposed, self- and mutual-inductances as
well as series and shunt capacitances were studied in detail. Inductance and capacitance
variation due to axial and radial deformations was discussed analytically. A numerical
example was presented and it showed that the inductance value will be changed due the
axial deformation; this alteration for capacitance was insignificant. Study on radial
deformation revealed the shunt capacitance variation in transformer winding. This in turn
almost verifies experimental findings in ‎[102]. It should be emphasized that different
winding deformations can lead to different inductance and capacitance variation and in a
complex deformation; these might not even be distinguishable. Therefore, having obtained
the frequency response spectrum, it is suggested to carry out calculations using the
analytical approach prior to FRA interpretation. However, deformations with similar
parameter changes lead to similar deviation pattern in the FRA spectrum, viz., inductance
reduction either in a single winding or a single phase transformer results in identical
patterns (not identical deviation) in FRA trace movement.
112
6.6 FRA Simulation Study and Practical Measurement Results
Transformer winding modelling to interpret oscillations was discussed in detail in Chapter

4. In addition, axial and radial deformations of the transformer winding and their
influences on winding parameters were investigated in this Chapter. The current Section
tries to explore the FRA trace deviation due to each and every parameter variation in
transformer winding. To gain this knowledge, the mathematical model presented in
Chapter 4 was used. Transformer winding inductance, series and shunt capacitances, turn-
to-turn and turn-to-ground conductance are changed and their impacts on the FRA trace
are discussed. Afterwards, the simulation results are compared to practical measurements
to verify our approach in the interpretation of FRA spectrum due to winding deformation.
It should be noted that due to the high costs associated with a transformer or perhaps
winding, the experiments associated with this research had to be non-destructive.
Therefore in order to study all parameters to obtain the FRA data associated with different
levels of parameter changes, it was decided to utilise different methods which could
emulate inductance, capacitance, resistance and conductance variations. In addition,
different test objects are used to demonstrate that fundamental parameters variation in
transformer winding will lead to similar deviation pattern in the FRA spectrum. Hence, the
simulations are carried out on the unique model transformer and verification is performed
through different test objects.
6.6.1 Inductance Variation (Simulation)
One of those crucial parameters in transformer winding is the winding inductance.

Winding inductance contains self-and mutual-inductance as discussed in (3.1). It is mostly
influenced by winding configuration, magnetic core reluctance or even active part
condition. Any changes of these factors can alter the self- or mutual-inductance and
ultimately the total winding inductance. In practice, the transformer winding self-
inductance cannot be increased easily. In fact, it is tightly related to the transformer
magnetic reluctance, see equation (5.8). It was discussed in Chapter 5 that typical designs
by transformer manufacturers aim to minimize transformer magnetic reluctance. Hence,
any deformation or undesirable lamination of transformer core sheets will lead to increase
in the magnetic reluctance and accordingly, the winding self-inductance will have lower
value. The other parameter which influences inductance is the number of winding turns. If
the number of winding turns is increased, the winding inductance would be increased as
well, but this obviously is not possible practically.
113
On the other hand, this Chapter revealed analytically that the axial deformation of winding
will lead to a decrease in winding inductance. This was further demonstrated by a
numerical example. Hence, decreasing of the winding inductance is more feasible in
practice as compared to increasing Based on this, it was supposed that the total
inductance of the winding has decreased for 20% and 40% from the reference value, and
the simulation was carried out on the mathematical model of winding presented in
Chapter 4. Figure 6.9 shows the simulation result.
20
-20
Magnitude [dB]
-40
-60
-80 Frequency Band : 1 kHz - 20 MHz

L
0.8 x L
0.6 x L
-100 4 5 6 7
10 10 10 10
Frequency [Hz]
(a)
20
L
10 0.8 x L
0.6 x L
0 Frequency Band : 1 MHz - 10 MHz
-10
-20
Magnitude [dB]
-30
-40
-50
-60
-70
-80
-90
6 7
10 10
Frequency [Hz]
(b)
Figure 6.9. FRA simulation results of winding due to the inductance reduction for 20 and 40 %, (a) Entire FRA
spectrum, (b) Expanded view of dash-line rectangle region in Fig. 6.9(a).
114
According to Fig. 6.9, self-inductance reduction will lead to shifting of the first anti-
resonance in transformer FRA trace to higher frequencies. Therefore, deviation of FRA
spectrum toward the right hand side in the low-frequency band can be interpreted
through self-inductance reduction. However, this movement in FRA trace could be affected
through the shunt capacitance reduction as well. The shunt capacitance will be discussed
in following subsections. In addition, the anti-resonances in Fig. 6.9(b) have reduced in
magnitude. This can be interpreted through smaller inductive impedance in (5.9) and so
less magnitude in (2.7) and/or (5.4). Therefore, inductance reduction in transformer
winding will cause anti-resonances shifting to higher frequencies and having smaller
magnitude. This deviation is more significant for the first anti-resonance in the FRA
spectrum. This is verified through practical measurement in the next subsection.
6.6.2 Inductance Variation (Practical Study)
In order to validate the simulation result, an 11/0.25 kV, 25 kVA transformer was used as a
test object and inductance reduction of the HV winding was studied. Technical
specifications of this transformer are given in Table 6.2.
Table 6.2. Technical specifications of single phase transformer.
Manufacture year 2010 Frequency [Hz] 50

Rated voltage [kV] 11/0.25 Number of phases 1
Rated power [kVA] 25 Cooling ONAN
At first, the frequency response trace was recorded from the HV side while the LV side was
left open circuit. This measurement configuration is considered a common setup for FRA
measurement (end-to-end measurement, see Fig. 2.10(a)). To study the inductance
reduction, the LV winding is deliberately short-circuited and the frequency response of HV
winding is recorded.
According to Section 5.2.2 and also Lenz’s law, making short circuit across the LV winding
will cause the magnetic flux initiated in HV winding unable to flow through the
transformer core. This in turn leads to significant increase in magnetic core reluctance and
the winding inductance is decreased. The FRA test setups for normal and reduced winding
inductances are illustrated in Figures 6.10(a) and 6.10(b). The measured result is shown
in Fig. 6.11.
According to Fig. 6.11, when the LV winding is short-circuited, the frequency response of
the HV winding changes in the range from 20 Hz to 40 kHz. For frequencies above 40 kHz,
115
there are no significant discrepancies between the recorded traces. The reason lies in the
fact that the self-inductance of the winding has changed due to the deliberately created
short circuit.
(a) (b)
Figure 6.10. FRA test setup on HV side with (a) Open circuit on LV side, (b) Short circuit on LV side.
In addition according to Fig. 6.11, the frequency response magnitude when the LV winding
is left open circuit starts from -60.78 dB and following its descending trend reaches the
minimal peak at 605.7 Hz. Also, the frequency response magnitude when the LV winding is
short-circuited starts from -8.99 dB at 20 Hz and following its descending trend reaches
minimal peak at 7.971 kHz, meaning that the frequency response trace has shifted to the
right considerably.
-10
-20
-30
-40
Magnitude [dB]
-50
-60
-70
-80
-90
-100 HV winding - LV (open circuit)
HV winding - LV (short-circuited)
-110 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 6.11. Frequency response of HV winding for open circuit and short-circuited LV winding.
Hence in low frequency range, the trace has shifted upward and toward the right
considerably. In order to interpret the discrepancy between traces, equation (2.7) was
116
taken into consideration. If assuming that inductance and capacitance are frequency
invariant, to calculate the self-inductance variation in open and short circuit situations,
(6.21) and (6.22) are employed:
 Vout (50 Hz ) 
20log10    11.9 dB (6.21)
 Vin (50 Hz ) 
 
 Vout (50 Hz ) 
20log10    62.33 dB (6.22)
 Vin (50 Hz ) 
 
According to Fig. 6.11, at the fundamental frequency (50 Hz), the frequency response
magnitude when the LV winding is short-circuited reaches -11.9 dB, while for open circuit
LV winding, it reaches –62.33 dB. According to (6.21) and (6.22), the impedance ratio for
open and short circuit situations is calculated as (6.23):
ZOC (50 Hz )  Zout 0.2540

 (6.23)
Z SC (50 Hz )  Z out 7.647 104
where, ZOC(50 Hz) is the inductive reactance of the HV winding in open circuit state at 50 Hz
and ZSC(50 Hz) is the inductive reactance of the HV winding at the same frequency when the
LV winding is short-circuited. Therefore, the ratio of open circuit and short circuit self-
inductance at the fundamental frequency for the tested transformer is given by:
LOC (50 Hz )
445 (6.24)
LSC (50 Hz )
where, LOC is the open circuit self-inductance and LSC represents short circuit self-
inductance. Thus, the open circuit self-inductance of this single phase transformer is 445
times greater than self-inductance of the HV winding when the LV winding is short-
circuited. Since the number of turns of the HV winding has not changed, the calculated
ratio for inductances can be applied to reluctances accordingly. This experimental result
verifies the simulation study in the previous subsection.
6.6.3 Shunt Capacitance Variation (Simulation)
In order to explore the shunt capacitance variation on FRA spectrum, the total shunt
capacitance of the winding was decreased for 20% and 40% from the reference value, and
117
the simulation was carried out on the mathematical winding model presented in Chapter
4. Figure 6.12 shows the simulation results due to the shunt capacitance variation.
20
II
-20
L
Magnitude [dB]
I'" h
·I
-40
l!l li
~~ I
I'1: I'
liI !!
-60
-80
H~=~~.---------------~LI----
~ ---+-.~-----~-=·-=-,_..!/!
. ,~-'0)j
~\-_p·=~jl)'
Frequency Band : 1 kHz - 20 MHz
C (shunt)
-100 0.8 x C (shunt)
0.6 x C (shunt)
4 5 6 7
10 10 10 10
Frequency [Hz]
(a)
20
1=·==·=·---------------;
C (shunt)
I !
i -----· 0.8 x C (shunt) I
0
l::=:=~---------------1
0.6 x C (shunt)
-20
Magnitude [dB]
-40
-60
-80
-100
6 7
10 10
Frequency [Hz]
(b)
Figure 6.12. FRA simulation results of winding due to the shunt capacitance reduction of 20 and 40 %, (a)
Entire FRA spectrum, (b) Expanded view of dashed-line rectangle in Fig.6.12 (a).
118
According to Fig. 6.12, resonances and anti-resonances in FRA trace have moved to higher
frequencies and anti-resonance magnitudes have increased.
Resonant movements to the right hand side in the Bode diagram can be interpreted
through equation (5.20), while magnitude increase in FRA trace comes through interaction
between capacitive and inductive reactances. Since the shunt capacitance has decreased,
the resonant frequency has shifted to higher frequencies, and in the meantime, the
inductive reactance experiences higher amplitude due to the higher frequency (see
equation 5.20). Therefore, greater capacitive reactance is required to interact with
inductive reactance and creating an anti-resonance point in FRA trace.
6.6.4 Shunt Capacitance Variation (Practical Study)
In order to study the shunt capacitance variation, the single phase transformer introduced
in Table 6.2 was again taken as the test object. To change the shunt capacitance, the test
setup was configured as in Fig. 6.13 where the LV winding was short-circuited and
transformer tank was isolated from the ground. Note that the transformer tank was
originally grounded (see Fig. 6.10(b)).
Figure 6.13. FRA test setup for HV winding, where LV side is short-circuited, and test object tank is isolated
from the ground.
Obviously from a physical point of view:
CLC , CHL and CHT  0 (6.25)
where CLC, CHL and CHT denote shunt capacitances of LV to the core, HV the LV and HV to the
transformer tank, respectively. For the shunt capacitance configuration in Fig. 6.14:
C C C C
Csh  CHT  HL LC 
 Csh  HL LC
(6.26)
( HV ) C C ( HV ) C C
HL LC HL LC
where, Csh(HV) and Cʹsh(HV) are the HV winding shunt capacitances when the tank is grounded
and isolated, respectively. Equation (6.26) is always true. Therefore, we expect to get a lower
119
value of the shunt capacitance for HV winding through isolating the transformer tank. This
hypothesis was verified by FRA measurement on the HV side of single phase transformer for
the test setups in Figures 6.10(b) and 6.13. The results are shown in Fig.6.15.
HV HV
CHT CHL CHT CHL

LV LV
Tank Tank
CLC CLC
Core
HV HV
CHT CHL CHT CHL

Csh(HV) LV C’sh(HV)
CLC
(a) (b)
Figure 6.14. Active part and related shunt capacitances, (a) Shunt capacitance configuration for HV, where LV
side is short-circuited and transformer tank is grounded, (b) Shunt capacitance configuration for HV, where
LV side is short-circuited and transformer tank is isolated.
-10
-20
-30
Magnitude [dB]
-40
-50
-60
-70 9.003 kHz

-69.16 dB
Frequency Band: 20 Hz - 2MHz
Shortened LV side 7.883 kHz

-80 -68.7 dB
Shortened LV side - Tank Isolated
2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 6.15. Frequency response of HV winding of single phase transformer, LV winding is short-circuited,
transformer tank is grounded and isolated.
120
According to Fig. 6.15, the frequency response trace for grounded tank does not deviate
from the isolated one over the very low frequency range. In fact, the self-inductance of the
HV winding does not change for grounded and isolated tank.
Moving from very low to low frequencies reveals that the minimal peak has shifted to a
higher frequency (from 7.883 kHz to 9.003 kHz) and its absolute magnitude has increased
from 68.7 dB to 69.16 dB. In addition, in mid and high frequencies the trace of isolated
tank displays greater absolute magnitude compared to grounded tank, while oscillation
trends in both traces remain roughly the same.
To interpret mid and high frequency discrepancies in Fig. 6.15, it is worth noting that the
first minimal peak in FRA trace can be calculated as (5.20). Since HV winding configuration
has not changed and just the tank has been isolated, the series capacitance of HV winding
has not varied. Based on this fact, minimal peak movement in Fig. 6.15 from 7.883 kHz to
9.003 kHz can be interpreted as due to winding shunt capacitance variations. In fact, anti-
resonance movement in the trace of isolated tank to higher frequency implies a slight
reduction in shunt capacitance of the winding. This showed in (6.26) and in turn verifies
the simulation results for the case of shunt capacitance reduction in subsection 6.6.3.
On the other hand, capacitive reactance increases when capacitance of the winding
decreases. This in turn results in greater absolute frequency response magnitude in mid
and high frequencies.
Note that, the self-inductance does not change during isolated and grounded tank while
the shunt capacitance has decreased.
6.6.5 Series Capacitance Variation (Simulation)
The series capacitance can be changed due to changes of turn-to-turn or disk-to-disk

distances of transformer winding. Any change or movement of the conductor or slanting of
the vertical form can also influence the winding series capacitance. Tilting, bending and
conductor deformation can change this parameter as well. In order to examine the impact
of series capacitance of the winding on FRA trace, this parameter was increased by 20%
and 40% and simulation was performed to obtain the corresponding FRA traces. The
result showed that such changes in series capacitance did not influence the FRA trace
significantly. Hence, changes of series capacitance were examined for larger values. It was
supposed that the series capacitance of the winding experiences 100 times and 200 times
greater than the original value. This could be happened when the transformer winding
121
turns are tilted. Figure 6.16 shows the FRA simulation result for the series capacitance
variation.
20 t
-20 ~
~
I
Magnitude [dB]
I
!
-40
~
I
I
' I
I
I
I I
I
-60
i
"
~
II
ll
J\
1/1
I
II'
-80 Frequency Band : 4 kHz - 20 MHz
r---------------------1
1· · - · C (series)
II -----
l-----
100 x C (series)
200 x C (series)
-T-- 1
I
I
I
_l
~ v \j ~~
-100 4 5 6 7
10 10 10 10
Frequency [Hz]
(a)
20 -~--- - - -~--~1
i·-- C (series) I
! ----· 100 x C (series) !
200 x C (series)
L-=--=.-=--=------------------J
0
I
I I
I
I
I
I
-20 I '
iI
Magnitude [dB]
I I
! ~ ..
I I'
-40
I 1\i
Ill
'
iit !
-60 1 il\ i
r-----__ L i ""-/1 iv ~ ~\·y \

I'''
: I(
/i I
1,1
I· I
11 \
'/ l
------
-80
/ ~ I \
j _l_
-100 6 7
10 10
Frequency [Hz]
(b)
122
20
13.770 MHz
C (series) 13.990 MHz
100 x C (series)
200 x C (series)
0
Frequency Band : 4 MHz - 20 MHz
13.560 MHz
-20
Magnitude [dB]
-40
-60
-80
-100 7
10
Frequency [Hz]
(c)
Figure 6.16. FRA simulation results of winding due to the series capacitance increase, (a) Entire FRA spectrum,
(b) Close-up view of dash-line rectangle in Fig. 6.16 (a), (c) Close-up view including some resonance
frequencies.
According to Fig. 6.16, series capacitance variation will influence the mid frequencies
towards the high-frequency band in the FRA spectrum. In fact, it has insignificant impact
on the first anti-resonance in FRA trace. In addition, as can be seen in Fig. 6.16, series
capacitance alteration of the transformer winding does not influence the low-frequency
band. Also, it tends to change the higher frequencies more than the lower frequencies.
Indeed, series capacitance variation influences the FRA spectrum less than shunt
capacitance. In other words, the shunt capacitance has more impact on the FRA spectrum
than the series capacitance.
6.6.6 Series Capacitance Variation (Practical Study)
A transformer had been de-energized by emergency disconnection system actuated by

gassing and differential protection systems, as well as two safety valves. Thus, in order to a
detailed investigation of the transformer prior to re-energizing, various diagnostic tests
were conducted. One of those tests was FRA measurement on different phases of the faulty
transformer. Fig. 6.17 shows the FRA result for phases b and c of the LV side. According to
this, the transformer was suspected to have winding deformation. Hence, the transformer
oil was drained and transformer tank was removed for internal inspection. Visual
inspection showed that the winding conductors of phase c in LV side have tilted. It was
123
highlighted earlier that conductor titling can in turn cause the series capacitance to change
significantly. In practice, the FRA trace has been deviated in mid frequencies for phase c
due to the titling, while the low and high frequencies are following normal oscillations (see
Fig. 6.17).
-5
-10
-15
Magnitude [dB]
-20
-25
-30
-35
-40
-45 ~ ---· Phase b (LV side)
I
Phase_ c (LV side)

L-:_-:_-:_·______
-50 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
(a)
-5
-10
-15
Magnitude [dB]
-20
-25
-30
-35
-40
Frequency Band: 15 kHz - 1.1 MHz
-45 ~ ---· Phase b (LV side) I
: Phase c (LV side) l
l-:_-:_-:_·______
-50
::::::::::::::::::::J:::::c___::::___L::::L_______________
c _ _______ ._ _ _ _ _ _c __ __ ._ __ ._ __ .____ _
5 6
10 10
Frequency [Hz]
(b)
Figure 6.17. FRA measurement results of faulty transformer for phase b and c, (a) Entire FRA spectrum, (b)
Close-up view of dash-line rectangle in Fig. 6.17 (a).
124
Figure 6.17(b) provides the close view of the FRA traces. According to simulation results,
changes in series capacitance will affect the mid-frequency band of FRA trace; practical
results illustrated in Fig. 6.17 verify the simulation results.
6.6.7 Resistance Variation (Simulation)
In order to study the impact of winding resistance (r) variation on the FRA spectrum, this
parameter was changed for 5 and 10 %, respectively, and the frequency response was
simulated for the model winding. Figure 6.18 shows the simulation results. According to
Fig. 6.18(a), variation in winding resistance can influence the very low and also very high
frequencies. Figures 6.18(b), (c) and (d) show the close-up views of those areas enclosed
by dash-line in 6.18(a). Other parts of the spectrum remain intact.
10
-10
c
b
-20
-30
Magnitude [dB]
-40
-50
-60
-70
Frequency Band : 20 Hz - 20 MHz d
-80 r
1.05 x r
-90 1.1 x r
3 4 5 6 7
10 10 10 10 10
Frequency [HZ]
(a)
125
-2
-4
Magnitude [dB]
-6
-8
-10
-12
-14 Frequency Band : 20 Hz - 45 kHz
r
-16 1.05 x r
1.1 x r
-18 3 4
10 10
Frequency [HZ]
(b)
5
r
1.05 x r
0 1.1 x r
Frequency Band : 18.61 MHz - 18.68 MHz
-5
Magnitude [dB]
-10
-15
-20
-25 7.27 7.271

10 10
Frequency [Hz]
(c)
126
-85.329
-85.3292
-85.3294
Magnitude [dB]
-85.3296
-85.3298
-85.33
Frequency Band : 19.99 MHz - 20 MHz

-85.3302 r
1.05 x r
-85.3304 1.1 x r
7.30101 7.30101
10 10
Frequency [Hz]
(d)
Figure 6.18. FRA simulation results of winding due to winding resistance increment, (a) Entire FRA spectrum,
(b) Close-up view of region ‘b’ , (c) Close-up view of region ‘c’ , (d) Close-up view of region ‘d’.
At very low frequencies, the capacitive reactance due to the total capacitance of
transformer winding shows value high enough to be considered as open circuit. Therefore,
the winding impedance in this frequency range is more influenced by the inductive
reactance and winding resistance.
As the frequency increases, the magnitude of the inductive reactance becomes

considerable and thus winding resistance impact becomes negligible. Hence, winding
behaviour follows the inductive reactance trend. This occurred at 10 kHz in Fig. 6.18. In
fact, before 10 kHz the winding resistance value and its changes are significant as
compared to the winding inductive reactance. However, this is not so after 10 kHz.
At very high frequencies, the inductive reactance due to the total inductance of winding is
large enough to be considered as open circuit. Consequently, the capacitive reactance due
to total winding capacitance and the winding resistance are more dominant in FRA trace.
However, the capacitive reactance experiences small value as the frequency is significant
in this region. Therefore, the winding resistance and its variation become more significant
as depicted in Fig. 6.18.
6.6.8 Resistance Variation (Practical Measurement)
In order examine the winding resistance variation practically; the glassy transformer
model was taken as a test object. The AC resistance of the HV winding was measured and a
127
resistor having 5 % value of the total HV winding resistance was connected to the winding
(series combination). This setup can model the winding resistance increment. Afterwards,
the frequency response of HV winding was re-measured. Similar experiment was
performed for the 10 % increment in the total winding resistance value using the same
approach. The results of measured frequency response are shown in Fig. 6.19.
According to Fig.6.19, practical measurements show a trend similar to what happened for
simulation results. In fact, very low and also very high frequencies have been influenced
and deviated due to the winding resistance variation. Comparing simulation and
measurement results, it is clear that measurement result verifies simulation result.
b
-10
c
-20
Magnitude [dB]
-30
d
-40
-50
Frequency Band: 2 kHz - 20 MHz
r
-60 ~1.05 x r
~1.1 x r
4 5 6 7
10 10 10 10
Frequency [Hz]
(a)
128
-1
-2
Magnitude [dB]
-3
-4
-5
Frequency Band: 2 kHz - 4.5 kHz

-6 r
~1.05 x r
~1.1 x r
-7
4
10
Frequency [Hz]
(b)
-18.5 r
~1.05 x r
-19 ~1.1 x r
Frequency Band: 14.60 MHz - 16.82 MHz
-19.5
-20
Magnitude [dB]
-20.5
-21
-21.5
-22
-22.5
-23
-23.5
7.17 7.18 7.19 7.2 7.21 7.22
10 10 10 10 10 10
Frequency [Hz]
(c)
129
-20
-22
-24
Magnitude [dB]
-26
-28
Frequency Band: 17.5 MHz - 20 MHz

-30
r
~1.05 x r
~1.1 x r
-32
7.25 7.26 7.27 7.28 7.29 7.3 7.31
10 10 10 10 10 10 10
Frequency [Hz]
(d)
Figure 6.19. FRA measurement results of winding due to the resistance increment, (a) Entire FRA spectrum,
(b) Close-up view of region ‘b’ at very low frequencies, (c) Close-up view of region ‘c’ at high frequency
resonance, (d) Close-up view of region ‘d’ at very high frequencies.
6.6.9 Conductance to Ground (G) Variations (Simulation)
The insulation conductivity can be changed due changes of insulation parameters such as
the loss factor – also called the Dielectric Dissipation Factor (DDF). DDF is affected mainly
by temperature and moisture content of the insulation medium. Also, DDF in paper
insulation could be affected through the degree of polymerization (DP). In the case of oil
insulation, DDF might be influenced by oil contamination, viscosity, acidity, interfacial
tension, etc. All in all, increase in loss factor deteriorates the insulation quality and
changes the insulation conductance in the insulation model. The insulation system
becomes vulnerable due to significant conductance increment. As a rule of thumb, a
temperature increase of 10°C will halve the insulation resistivity and make insulation
conductivity two times greater ‎[105].
Based on this, the conductance to ground (G) of the transformer winding was changed and
the frequency response was simulated to realize its impact on FRA spectrum. Figure 6.20
shows the FRA simulation result, where the conductivity to ground for transformer
winding was changed to 5 and 10 times greater than the original value.
130
10
-10 ~
-20
""""
Magnitude [dB]
-30
-40
-50 ""
-60
-70 """' I
I
-80 I
I
I
-
-90 H ------- 10 x G
-
- G
L--------r------- -
5xG
10
4
I
!
I
l
10
5
" Frequency [Hz]
~
6
10
/
I' I \J J lJ
7
10
(a)
I J . _l l I I I
G
10 rl ==~~~=~~~:
".
5xG ~ 4.664 MHz, 8.367 dB
I
I -·-·-·-·•-·-·-·-··
L _____ 10 x G
0
-10
// \
~
I!'\ 4.664 MHz, -11.97 dB

-~
-20
I '\
Magnitude [dB]
-30
-40
/ \.
-50
-60
-70
-80
~
~
/
"' ~
:::::;:::
--
-90
6.61 6.63 6.65 6.67 6.69 6.71

10 10 10 10 10 10
Frequency [Hz]
(b)
Figure 6.20. FRA Simulation results of winding due to the conductance to ground (G) increment, (a) Entire FRA
spectrum, (b) Close-up view of the first resonance point.
131
According to the simulation, changes of conductance (G) in the winding model can
influence the frequency response since it modifies the dielectric permittivity and the
losses. This causes resonance damping in the FRA trace. For instance in the current case,
the magnitude of the resonant peak in Fig. 6.20 (b) was changed from 8.367 dB to -11.970
dB while the resonance frequency remained unchanged.
6.6.10 Turn-to-Turn Conductance (g) Variations (Simulation)
Similar to the last subsection, the turn-to-turn conductance (g) was also examined for
values 5 and 10 times larger than the original conductance and frequency responses were
simulated. Fig. 6.21 illustrates the simulation results.
10
-10
-20
Magnitude [dB]
-30
-40
-50
-60
-70
-80 g
5xg
-90 10 x g
4 5 6 7
10 10 10 10
Frequency [Hz]
(a)
132
4.664 MHz, 8.362 dB
5
4.664 MHz, 3.737 dB
0
Magnitude [dB]
-5
-10
-15
-20 g
5xg
10 x g
6.66 6.67 6.68
10 10 10
Frequency [Hz]
(b)
Figure 6.21. FRA Simulation results of winding due to the turn-to-turn conductance (g) increment, (a) Entire
FRA spectrum, (b) Close-up view of the first resonance point.
According to Fig. 6.21, similar to (G), the change of turn-to-turn conductance (g) has
influenced the frequency response, since it modifies dielectric permittivity and the losses.
This in turn causes a slight magnitude reduction at the resonance points in the FRA trace.
The magnitude deviation in Fig.6.20 due to the conductance (G) variation was 20.337 dB,
while similar experiment for (g) led to 4.624 dB deviation in FRA spectrum magnitude.
The former is almost 5 times greater than the latter. This in turn reveals that FRA trace is
more influenced by the variation of conductance to ground (G) rather than turn-to-turn
conductance (g) alteration.
6.6.11 Conductance Variation (Practical Study)
Practical study on turn-to-turn (g) and turn-to-ground (G) conductance will be discussed
in detail in Chapter 7, where the influence of moisture and temperature changes is going
to be studied.
6.7 Conclusion
Study on axial and radial deformation was conducted analytically in this Chapter. It was
addressed that the transformer winding inductance will be changed due to the axial
deformation to winding disk. In addition, radial deformation will change the shunt
capacitance of the winding significantly. The analytical approach was completed through a
133
numerical example as well as simulation study using finite element method. A practical
study was also conducted to verify the analytical approach.
This Chapter also paid attention to the FRA trace deviation due to the winding parameters
changes. Simulation study was performed on the winding model presented in Chapter 4
and practical results were achieved on different test objects. Practical results have verified
the simulation studies. In summary, the influences of winding inductance, series and shunt
capacitances, resistance, conductance to ground and turn-to-turn conductance on FRA
trace were clarified in this Chapter.
134
Chapter 7. Temperature and Moisture Content Influences on FRA Signature
Chapter 7
Temperature and Moisture Content

Influences on FRA Signature
7.1 Introduction
This Chapter is focused on the influence of temperature and moisture migration on the
FRA trace of transformer winding. It also discusses the feasibility of FRA capability in
moisture diffusion recognition in transformer paper insulation. To conduct this
investigation, the manufactured glassy model transformer and a 20/0.4 kV, 1.6 MVA three-
phase two windings transformer are used as test objects in the experiment.
At first, transformer water dynamic is discussed. After that frequency response

measurements on HV and LV windings of the test object at different moisture contents for
various temperatures are performed and deviation in resonance frequencies discussed.
Total capacitance deviation of transformer windings due to the moisture and temperature
changes is calculated using mathematical approach. In the meantime, Karl-Fischer
Titration (KFT) is utilized as a method to measure the Water Content of the Oil insulation
(WCO), and the Water Content of the Paper insulation (WCP) is then derived using MIT
equilibrium curve. In addition, Dielectric Dissipation Factor (DDF) is measured for each
and every stage. The result achieved through the model transformer is then examined on a
real (power-rated) transformer. Based on practical results, a hypothesis is proposed for
“the main reason of FRA trace deviation due to the moisture and temperature variation”.
This hypothesis is verified through simulation result. Finally, FRA statistical indicators are
calculated and discussed in detail, and possible solutions to distinguish insulation
characteristic impacts on FRA trace from winding deformation is provided.
7.2 Transformer Water Dynamic
Residual moisture in a transformer due to the water ingress through atmosphere,

insulation aging, cellulose decomposition or even after dry-out process will transfer from
135
the oil into the paper insulation and from the paper towards oil insulation in low and high
temperatures, respectively. Indeed moisture migration from one phase (liquid/solid) to
the other phase (solid/ liquid) could be due to the moisture concentration, temperature
and pressure gradients ‎[106]. Nevertheless, water dynamics in transformer can be
classified into transient and steady state periods ‎[107]. The transient period involves
moisture diffusion activity whereas water equilibrium between paper and oil insulations
is attained in the steady state. Fig. 7.1 shows the water dynamic in paper and oil
insulations for different temperatures.
Figure 7.1. Water dynamic in paper and oil insulation for different temperatures T1 and T2 (T1 < T2), WCO and
WCP; t denotes the time.
7.2.1 Transient
In the case of moisture diffusion, Fick’s second law as one of the basis equations can be
expressed as:
C  2C
c D c (7.1)
t x2
where D is the diffusion coefficient [m2/s], Cc is the concentration of substance (moisture)

[mol/m3] and x denotes substance movement position [m].
Temperature changes can lead to different activation energy for the molecules in adjacent
regions and ultimately moisture migration. Guidi and Fullerton have employed a diffusion
model to estimate moisture migration from the transformer paper insulation ‎[52], ‎[106]
and ‎[108]. They have specifically focused on power transformer drying time, temperature
and moisture adsorption rates for insulation when exposed to the atmosphere ‎[106]. Their
estimation is expressed by:
136
  
 k c  E  1  1 
 0 T T  (7.2)
D  D0e   0 k 
where c is the local moisture concentration [H2O/kg], Tk is the current temperature [°K], T0
is the reference temperature (298 °K), kʹ is a dimensionless parameter and is equal to 0.5
‎[109]-‎[110], D0 is a pre-exponential factor [m2/s] and E0 is the activation energy of the
diffusion process [kJ.mol-1]. D0 and E0 have been estimated for oil-free as 2.62×10-11 and
8140 and for oil-impregnated paper as 1.34×10-13 and 8074 by ‎[109]-‎[110], respectively.
Substitution of (7.2) into (7.1) will lead to (7.3) which can explain moisture migration as a
function of temperature. The solution of (7.3) is detailed in ‎[111]:
  
 0.5c 8140 1  1   2
C 
 T   C
 0 Tk   (7.3)
c  2.62 10 e 
11 c
t x2
7.2.2 Equilibrium
Water content equilibrium between paper and oil insulation has been widely discussed in
‎[110], ‎[112]-‎[115]. Equilibrium curves showing paper water content versus oil water
content for different temperatures have been achieved; hence, it is possible to determine
the value of one of them once knowing the other ‎[112]. Most of the above-mentioned
literatures have provided the equilibrium curves up to at most 100 ppm, a wider range of
equilibrium curve from 0 °C to 100 °C and moisture in oil up to 800 ppm was presented in
a work by Du et al at MIT university and named MIT oil-paper equilibrium curve ‎[110].
Therefore, due to the wide range the MIT equilibrium curves were employed in the
current study to derive the moisture content.
The diffusion time constant for moisture diffusing from one side of insulation is also
calculated as ‎[110]:
d 2p
 4 , (7.4)
D 2
While, this equation for double-side diffusion through paper insulation is given by ‎[110]:
d 2p
 . (7.5)
D 2
where τ is the diffusion time constant, and d represents the thickness of paper insulation
or pressboard.
137
7.3 Practical Study
7.3.1 Test Object and Setup
To conduct an accurate practical study on temperature and moisture influences on FRA

spectrum of winding, the glassy model transformer was used in the experiment. Detailed
information on this test object is provided in Appendix D as well as Chapter 4.
The test object is shown in Fig. 7.2. Transformer oil is required to be injected into the
glassy container. Hence, a drain valve was fitted onto the top plate to enable oil injection
and also taking oil sample.
Figure 7.2. Manufactured glassy air-core transformer (setup preparation to study temperature and moisture
impact).
Three different methods can be used for heating the test object: Low Frequency Heating
(LFH), oil circulation through an oil circulator, and using an electric oven. Among these,
the last approach was chosen for this study as it would be more accurate for controlling
the temperature. The test object was placed inside the oven and wiring connections for
FRA measurement were brought out through a bushing mounted on a small opening on
the oven top. The oven was equipped with a sensitive thermostat and a digitally controlled
heater to govern the internal temperature. Two thermocouples were used to monitor the
138
internal temperature of the oven as well as the test object. The FRA test setup then
remained unchanged for the entire experiment. FRA measurement was performed on a
winding by injecting a swept sinusoidal signal Vin at the line-lead, and detecting the
response Vout at the neutral-lead as shown in Fig. 7.3.
Figure 7.3. FRA test setup to examine temperature and moisture variation.
As discussed in Chapter 2, FRA measurement is performed in the frequency band 20Hz –

1MHz for transformers with highest voltage of > 72.5 kV, and in the range of 20Hz – 2MHz
for transformers with highest voltage of ≤ 72.5 kV ‎[1]. To be on the safe side, FRA
measurement is performed at least over the range 20Hz – 2MHz for all transformers
irrespective of their voltage rating. However, in the case of special transformers or
reactors, the upper limit may be shifted to higher frequencies. For instance with air-core
reactors, this limit could be even increased up to 20 MHz. In this study, the significant
value of the air-core magnetic reluctance of the test object results in a small self-
inductance for the windings. Small self-inductances will lead to small inductive reactances.
Therefore, the resonance frequencies in FRA traces for HV and LV windings would be
shifted considerably to higher frequencies. Hence, the upper band limits for FRA
measurements were extended from 2 MHz to 20 MHz to display entire oscillations similar
to what conducted for FRA measurement in Chapter 4.
7.3.2 Case Study 1 (‘Wet’ Model Transformer)
To study the temperature and moisture variation on FRA trace, at first the ‘wet’ test object
was examined. Initially, the transformer drain valve was opened and the test object was
deliberately left exposed to the laboratory ambient for two weeks so that the paper
insulation was saturated with moisture. The average readings of ambient temperature and
139
relative humidity were 23 °C and 26 %, respectively. Hence, the initial moisture content
for paper insulations was 4.1 %, calculated using the following data in Table 7.1 and also
equation (7.6) provided by Du et al ‎[109] on air relative humidity method.
Table 7.1. Ambient air relative humidity and moisture in paper (polynomial fitting parameters for various
temperatures), taken and modified ‎[109].
30 °C 40 °C 50 °C 60 °C 70 °C
a0 (x 10-1) 2.4131270 1.6954583 1.0483257 1.3978572 0.7441865
a1 (x 10-1) 3.2828657 2.9079147 2.4316118 2.1359436 1.7762623
a2 (x 10-3) -14.929696 -11.950117 -7.2850779 -6.2300223 -2.7797731
a3 (x 10-4) 4.3831525 3.2448905 1.2731316 1.1731076 -0.27101029
a4 (x 10-6) -6.3395879 -4.2926236 -0.37397578 -0.57129397 2.2473555
a5 (x 10-8) 4.2446633 2.5228351 -1.1755019 -0.75286519 -3.3218692
a6 (x 10-11) -9.3468655 -3.8729882 9.5319144 7.4380470 16.252499
WCP  a6(RH )6  a5(RH )5  a4(RH )4  a3(RH )3  a2(RH )2  a1(RH )  a0 (7.6)
where WCP is the moisture in paper in percent by weight and RH is the air relative
humidity in percent.
Afterwards, the transformer oil was dried-out (see appendix E), and the glassy tank was
filled with dry transformer oil (< 5ppm, at 70 °C) then left until the paper is fully
impregnated by oil, and eventually both of them reached equilibrium. The equilibrium
time was calculated as 244 hours using (7.2) and (7.4) at 23 °C. Thus, the oil sample was
taken from the container using a glassy syringe after 11 days (>244 hours). This is long
enough to ascertain that oil and paper insulation are in equilibrium.
The moisture content of the oil was measured using KFT method at 23 °C and moisture
content of the paper insulation was derived through MIT oil-paper equilibrium curves
‎[110], accordingly. These values were 11 ppm and 4 %, respectively.
Derived moisture contents of the paper insulation through air relative humidity and KFT
methods led to similar results (4.1 % and 4 %).
Next, the test object was heated up to 30 °C to be ready for the initial stage of FRA
measurement. FRA measurements were performed over the range of 30 to 90 °C in 10 °C
increments but the spectra were recorded for HV and LV windings at 30, 50, 70 and 90 °C.
The frequency response trace for HV winding was recorded over the range 20 Hz – 20
MHz when the LV winding terminals were left open circuit (end-to-end measurement).
Likewise, FRA measurement was also performed for the LV winding. On completion, the
140
oven temperature was increased gradually to reach 50 °C. The test object was left standing
for 25 hours (equilibrium time at 50 °C) to complete heat exchange between the test
object insulations and oven environment and also to pass the equilibrium time.
-10
-20
-30
Magnitude [dB]
-40
-50
-60
-70
-80 HV winding spectrum (at 30 °C )
HV winding spectrum (at 90 °C ) Region 1 Region 2 Region 3
-90 4 5 6 7
10 10 10 10
Frequency [Hz]
(a)
0
2.623 MHz (at 30 °C)
1.742 MHz (at 50 °C)

-10 1.761 MHz (at 30 °C)
2.594 MHz (at 50 °C)
1.733 MHz (at 70 °C) 2.552 MHz (at 70 °C)
-20
2.540 MHz (at 90 °C)
1.705 MHz (at 90 °C)
-30
Magnitude [dB]
-40
-50 1.802 MHz (at 90 °C) 1.861 MHz (at 30 °C)
-60 1.832 MHz (at 70 °C)

1.841 MHz (at 50 °C)
-70
1.184 MHz (at 90 °C) HV winding spectrum (at 30 °C )
__ ..,....- ·-
HV winding spectrum (at 50 °C )
-80 1.190 MHz (at 70 °C) - --
-----fl
- l-- HV winding spectrum (at 70 °C )
1.209 MHz (at 50 °C) 1.223 MHz (at 30 °C)
-90 6
10
Frequency [Hz]
(b)
141
0
4.267 MHz (at 50 °C)
4.199 MHz (at 70 °C) 4.314 MHz (at 30 °C)
-10 4.178 MHz (at 90 °C)
7.668 MHz
-20 (at 30 °C)
7.344 MHz (at 90 °C)
-30 4.464 MHz (at 90 °C)
7.463 MHz (at 70 °C)
Magnitude [dB]
7.501 MHz (at 50 °C)
4.537 MHz (at 70 °C)

-40 4.610 MHz (at 30 °C) 8.863 MHz (at 90 °C)
4.560 MHz (at 50 °C)

-50 8.908 MHz (at 70 °C)
9.052 MHz (at 50 °C)

-60
9.052 MHz (at 30 °C)
-70
-90 7
10
Frequency [Hz]
(c)
Figure 7.4. FRA spectra for ‘wet’ model transformer, HV side, (a) Entire trace for 30 and 90 °C, (b) Close-up
view of region 1 shown by dash-line in Fig. 7.4(a), frequency band 800 kHz-3 MHz, (c) Close-up view of region
2 shown by dash-line in Fig. 7.4(a), frequency band 3 MHz – 10 MHz.
Then, frequency response traces for HV and LV windings were re-measured. Similar
experiments were done for 70 °C and 90 °C. FRA traces for HV and LV windings for
different temperatures are shown in Figures 7.4 and 7.5 over the frequency range of 5 kHz
– 20 MHz. The frequency response magnitude for the frequency range 20 Hz – 5 kHz was 0
dB and thus is not shown in Fig. 7.4 and Fig. 7.5. Figures 7.4(b), 7.4(c), 7.5(b) and 7.5(c)
show the close-up view of the measurement results.
Synchronous to FRA measurement, the moisture content of the oil insulation was
measured for all temperatures using KFT method. Moisture content recognition
techniques including KFT are described in Appendix F.
142
~'
I 1
,------
-10 I.
~
i
-20
J\~ i
! .
-( l
Magnitude [dB]
""' \ ~
-30
-40
"' \ j
I !
l
~
i
!
'!
I
I
-50 I
Frequency Band:..5 KHz
. . - 20 MHz
++ + + + + + + +
-- ---- ----
!-·-·· LV winding spectrum (at 30 °C) 1 I
!
!-· .. ·· LV winding spectrum (at 90 °C) ! Region 1 Region 2 Region 3
c____
L--· -·-· - ·r - ·-·-·- · + + + · - · - · - · · - · · -·-· - · .J
-60 4 5 6 7
10 10 10 10
Frequency [Hz]
(a)
0
LV winding spectrum (at 30 °C)
1.117 MHz (at 50 °C)
1.099 MHz (at 70 °C) LV winding spectrum (at 70 °C)
-10 LV winding spectrum (at 90 °C)
1.096 MHz (at 90 °C) 1.122 MHz (at 30 °C)
-20
Magnitude [dB]
-30
-40
0.928 MHz (at 90 °C)
-50 0.951 MHz (at 30 °C)
0.941 MHz (at 70 °C)

0.946 MHz (at 50 °C)
-60 6
10
Frequency [Hz]
(b)
143
0
6.817 MHz (at 70 °C) 6.855 MHz (at 50 °C)
4.380 MHz (at 70 °C)
6.722 MHz (at 90 °C) 6.963 MHz (at 30 °C)
4.404 MHz (at 50 °C)
-10
4.453 MHz (at 30 °C)
-20 4.367 MHz (at 90 °C)

Magnitude [dB]
-30 4.624 MHz (at 30 °C)

7.676 MHz (at 90 °C)
4.464 MHz (at 90 °C)
4.528 MHz (at 70 °C) 7.785 MHz (at 70 °C)

4.553 MHz (at 50 °C)
-40
7.915 MHz (at 50 °C)
-50 LV winding spectrum (at 30 °C) 8.039 MHz (at 30 °C)

-60 7
10
Frequency [Hz]
(c)
Figure 7.5. FRA spectra for ‘wet’ model transformer, LV side, (a) Entire trace for 30 and 90 °C, (b) Close-up
view of region 1 shown by dash-line in Fig. 7.5(a), frequency band 500 kHz - 3.5 MHz, (c) Close-up view of
region 2 shown by dash-line in Fig. 7.5(a), frequency band 3.5 MHz – 10 MHz.
Water content of the paper insulation was then derived through MIT equilibrium curves
for each and every temperature, independently.
The majority of the insulation system for manufactured test object was between HV and
LV windings including Kraft paper, oil canal, spacers and pressboard. Hence parallel to the
other measurements, the DDF values between HV and LV windings were measured at the
power frequency (50 Hz) for different temperatures to quantify the quality of insulation.
To avoid any kind of flash-over in the test object, the maximum applied voltage for DDF
measurement was kept low at 5 kV.
7.3.3 Discussion 1
According to Figures 7.4 and 7.5, when examining from low frequencies to 800 kHz all of
the traces are perfectly matched. Moving from 800 kHz to higher frequencies, the
discrepancy between the traces becomes obvious.
As the test object temperature changed from 30 to 90 °C, the FRA traces have slightly
shifted to lower frequencies. In the case of 90 °C, this movement seems to be more
significant. Some of the resonances and anti-resonances have been highlighted in Figures
7.4(b), 7.4(c), 7.5(b) and 7.5(c).
144
Detailed investigation of Figures 7.4 and 7.5 shows that the oscillations trend of the
spectra seems to be similar, while the resonance frequencies have shifted to lower
frequencies as the temperature increased and equilibrium time passed.
In the meantime, some of the resonant magnitudes are reduced. This deviation comes
certainly through windings’ inductance, total capacitance, resistance or insulation
conductance changes. From a mathematical point of view, resonances and anti-resonances
in FRA trace can be generated due to the interaction between inductive and capacitive
reactances. As frequency increases, each resonance point indicates the changing from
capacitive towards inductive behaviour, while the anti-resonance point shows the turning
from inductive to capacitive behaviour of the winding impedance. Hence, every resonance
or anti-resonance in FRA trace can be explained through (7.7) if it is considered
independently:
fi  1
2 L C (7.7)
i i
where fi is the ith resonance frequency, Li and Ci are the inductance and capacitance
(involving series and shunt capacitances) at ith resonance frequency. According to (7.7),
resonance frequencies in FRA trace would be changed if the inductance or total
capacitance is changed. Thus to clarify whether inductance variation due to the
temperature changes can influence FRA trace, the windings’ inductances were measured
for some frequencies and given in Tables 7.2 and 7.3 at 30 and 90 °C for HV and LV
windings, respectively. These tables also provide the measured values for windings’
resistances, inductive reactances and impedances.
According to Table 7.2 and Table 7.3, the maximum inductance deviation due to the
temperatures changes is less than 1.1 % for the entire measurement. Hence, it has
insignificant impact on FRA resonant peaks; therefore, deviation of resonant points in FRA
trace seems to be coming through changes in the total capacitance.
Total capacitance variation could be influenced by series and/or shunt capacitances

alteration.
145
Table 7.2. HV winding electrical parameters for 30 and 90 °C.
Parameter R X Z θ∡ L
100 Hz (30 °C) 0.087 Ω 0.444 Ω 0.453 Ω 78.91 707.90 μH
100 Hz (90 °C) 0.104 Ω 0.449 Ω 0.461 Ω 76.95 715.30 μH
Change 19.54 (%) 1.12 (%) 1.76 (%) 2.48 1.04 (%)
1 kHz (30 °C) 0.333 Ω 4.13 Ω 4.14 Ω 85.39 657.43 μH
1 kHz (90 °C) 0.347 Ω 4.17 Ω 4.19 Ω 85.24 664.57 μH
Change 4.20 (%) 0.96 (%) 1.20 (%) 0.17 1.08 (%)
100 kHz (30 °C) 11.65 Ω 337.76 Ω 337.95 Ω 88.02 537.55 μH
100 kHz (90 °C) 13.12 Ω 338.57 Ω 338.82 Ω 87.78 538.85 μH
Change 13.49 (%) 0.23 (%) 0.25 (%) 0.27 0.24 (%)
Table 7.3. LV winding electrical parameters for 30 and 90 °C.
Parameter R X Z θ∡ L
100 Hz (30 °C) 0.056 Ω 0.174 Ω 0.183 Ω 72.15 278.3 μH
100 Hz (90 °C) 0.067 Ω 0.176 Ω 0.188 Ω 69.15 280.3 μH
Change 19.64 (%) 1.14 (%) 2.73 (%) 4.15 0.71 (%)
1 kHz (30 °C) 0.114 Ω 1.67 Ω 1.68 Ω 86.09 267.25 μH
1 kHz (90 °C) 0.123 Ω 1.69 Ω 1.69 Ω 85.84 268.97 μH
Change 7.89 (%) 1.19 (%) 0.59 (%) 0.29 0.64 (%)
100 kHz (30 °C) 2.94 Ω 148.55 Ω 148.57 Ω 88.86 236.41 μH
100 kHz (90 °C) 3.33 Ω 148.93 Ω 148.97 Ω 88.72 237.03 μH
Change 13.26 (%) 0.25 (%) 0.27 (%) 0.15 0.26 (%)
Technically speaking, series and shunt capacitance can be altered due to the moisture
migration. As a hypothesis, the shunt capacitance is more influenced by transformer oil
insulation and thus it could be more significant in changing the total capacitance when
temperature and moisture change. This hypothesis is verified through simulation result in
Section 7.4.
Tables 7.4 and 7. 5 provide detailed information of deviated frequencies for FRA traces at
30 and 90 °C where f1 and f2 represent the identical frequencies of the anti-resonances and
resonances for HV and LV windings at 30 and 90 °C respectively. In addition, total
capacitance deviation of transformer windings for 30 and 90 °C have been calculated and
shown in the last column using (7.7).
According to Tables 7.4 and 7.5, average capacitance deviations for HV and LV windings
were calculated as 6.24 % and 6.20 %, respectively. Figure 7.6 shows the total capacitance
deviation of HV and LV windings for each 20 °C temperature increment. In Fig. 7.6, when
146
the temperature increases, the rising trend in capacitance deviation for both windings is
obvious due to the moisture migration from the paper insulation towards oil insulation.
HV 6.24 %
6 LV winding
LV 6.20 %
HV winding
5
LV 4.59 %
4 HV 4.12 %
Deviation (%)
3 LV 2.55 %
HV 2.43 %
2
20 30 40 50 60 70 80 90
Temperature ( °C )
Figure 7.6. Deviation of total capacitance for HV and LV windings (average deviation at 90 °C is 6.22 %).
Table 7.4. HV winding capacitance ratio, anti-resonance and resonance frequencies for quoted points in Fig.
7.4(b) and Fig. 7.4(c).
Frequency f1(30°C) f2(90°C) C2/C1=( f1/f2 )2 ΔC%

1st minimum 1.223 MHz 1.184 MHz 1.0669 6.7 %
1st maximum 1.761 MHz 1.705 MHz 1.0667 6.7 %
2nd minimum 1.861 MHz 1.802 MHz 1.0665 6.7 %
2nd maximum 2.623 MHz 2.540 MHz 1.0664 6.6 %
3 rd maximum 4.314 MHz 4.178 MHz 1.0661 6.6 %
3rd minimum 4.610 MHz 4.464 MHz 1.0664 6.6 %
4 th minimum 7.668 MHz 7.344 MHz 1.0901 9.0 %
6th minimum 10.810 MHz 10.580 MHz 1.0439 4.4 %
7th minimum 13.630 MHz 13.200 MHz 1.0662 6.6 %
4 th maximum 16.090 MHz 15.750 MHz 1.0436 4.4 %
The moisture content of the oil and paper insulations for each and every temperature was
recorded and shown in Fig. 7.7. Moisture migration from paper into the oil is clearly
obvious in this figure. The water content of the oil has increased from 18 ppm at 30 °C to
95 ppm at 90 °C.
147
Table 7.5. LV winding capacitance ratio, anti-resonance and resonance frequencies for quoted points in Fig.
7.5(b) and Fig. 7.5(c).

3 rd maximum 6.963 MHz 6.722 MHz 1.0729 7.2 %
100
Water content in oil (ppm) 95 ppm
WCP (%)
90 Water content in paper (%) 4.5
4%
80 3.8%
3.6%
70 3.5
63 ppm
3.0%
60
WCO (ppm)
52.5 ppm
2.6 %
50 2.5
2.2 %
2.0 %
40 1.8 %
1.7 %
35 ppm
30 1.5
21 ppm
18 ppm 25 ppm
20 15 ppm
10 13 ppm 0.5
0
20 30 40 50 60 70 80 90
Temperature ( °C )
Figure 7.7. Moisture content of oil and paper (wet model transformer).
In addition, the water content in paper insulation has decreased from 3.6 % at 30 °C to 1.7
% at 90 °C. In fact, changes of 1.9 % moisture content in paper insulation results in almost
6.22 % alteration in total capacitance of the windings at mid and high frequencies. This in
turn ultimately caused FRA spectrum deviation for HV and LV windings, respectively.
Additionally, based on Tables 7.2 and 7.3, the maximum deviation for the resistance of the
windings is 19.54 %, while this value is negligible for inductive reactance alteration. The
winding resistance variation can certainly influence the FRA trace magnitude in very low
frequencies, even though, when it comes to impedance and combine to the inductive
reactance its influence would be insignificant. Increasing the temperature will certainly
cause the dielectric loss in paper and oil insulations to increase. Dielectric loss variations
148
will marginally influence the magnitude of resonant peaks. Hence, some of the resonant
peaks in Figures 7.4 and 7.5 are only slightly damped due to the temperature increase.
Considerable dielectric loss will increase attenuation of the peaks. This in turn verifies the
simulation results on conductance variations in Chapter 6 (see subsections 6.6.9, 6.6.10
and 6.6.11).
Study on frequency response spectra when the temperature decreases from 90 to 30 °C

led to similar results. Although moisture absorption for paper insulation is different with
desorption, the equilibria should be the same and hence similar spectra were observed in
the reverse procedure.
The temperature and moisture influences on FRA trace for the ‘wet’ model transformer
(WCO 4%) were discussed. In order to realize the influence of initial moisture content
value on FRA spectrum deviation, the ‘dry’ model transformer is studied in the next
subsection and their results are compared.
7.3.4 Case Study 2 (‘Dry’ Model Transformer)
To study the impact of the temperature and moisture variation on FRA trace for dry
transformer winding, the oil was drained and the model transformer was heated up to 90
°C in the electric oven and then vacuumed (less than 750 mTorr) to remove the moisture
content from the paper insulation.
Figure 7.8 shows the vacuum process. After that, the vacuum was broken through a three-
way valve and then dry transformer oil was injected into the container once more. The
moisture content of the oil was measured after 11 days through KFT method and gave a
result of 3 ppm at 23 °C. The test object was heated up to 30 °C same as before to prepare
for frequency response measurement.
Figure 7.8. Vacuum process of the model transformer.
149
The FRA spectra for HV and LV windings were then measured at 30 and 90 °C over the
frequency range of 20 Hz – 20 MHz (end-to-end measurement), similar to the last
experiment. Also WCO, WCP and DDF at power frequency were measured as in the last
experiment. Figures 7.9 and 7.10 show the frequency response spectra just for 30 and 90
°C.
4.604 MHz (at 30 ºC)

-10 4.549 MHz (at 90 ºC)
-20
1.879 MHz (at 30 ºC)
1.857 MHz (at 90 ºC)
-30
Magnitude [dB]
-40
-50 1.986 MHz (at 30 ºC) 9.651 MHz (at 90 ºC)

9.768 MHz (at 30 ºC)
-60 1.963 MHz (at 90 ºC)
-70
Frequency Band: 900 kHz- 20 MHz
-80 1.305 MHz (at 30 ºC) HV winding spectrum (at 30 ºC), After dry-out
1.289 MHz (at 90 ºC) HV winding spectrum (at 90 ºC), After dry-out
-90 6 7
10 10
Frequency [Hz]
Figure 7.9. FRA spectra for ‘dry’ model transformer, HV side.
0
LV winding spectrum (at 30 ºC), After dry-out
LV winding spectrum (at 90 ºC), After dry-out 7.332 MHz (at 90 ºC) 7.407 MHz (at 30 ºC)
-10
1.194 MHz (at 30 ºC)
1.182 MHz (at 90 ºC)
-20
Magnitude [dB]
-30 4.865 MHz (at 30 ºC)
4.816 MHz (at 90 ºC)
-40
-50 1.011 MHz (at 30 ºC)
1.001 MHz (at 90 ºC)
-60 6 7
10 10
Frequency [Hz]
Figure 7.10. FRA spectra for ‘dry’ model transformer, LV side.
150
25
Water content in oil (ppm), After dry-out WCP (%)
21 ppm
Water content in paper (%), After dry-out
1.5 %
1.5
20 1.3 %
1.1 %
16 ppm
15
WCO (ppm)
1.0 %
0.9 % 1
0.8 %
10 10.3 ppm
0.6 % 0.5 %
8.0 ppm
0.5
0.4 %
5 5.0 ppm
3.0 ppm
3.0 ppm 3.0 ppm 3.0 ppm
0
20 30 40 50 60 70 80 90
Temperature (°C)
Figure 7.11. Moisture content of oil and paper (dry model transformer).
Moisture content migration in different temperatures in oil and paper insulations is

illustrated in Fig. 7.11. Measurements of electrical parameters for dry type study of HV and
LV windings resulted in values similar to those in Tables 7.2 and 7.3.
Table 7.6. HV winding capacitance ratio, anti-resonance and resonance frequencies for quoted points in Fig.
7.9.

Table 7.7. LV winding capacitance ratio, anti-resonance and resonance frequencies for quoted points in Fig.
7.10.

7.3.5 Discussion 2
It is noteworthy to mention that as a result of the dry-out process, all resonance

frequencies in Fig. 7.4 and 7.5 (‘wet’ model transformer) are shifted to higher frequencies
151
in Fig. 7.9 and Fig. 7.10 (‘dry’ model transformer), and the test object moisture content has
changed from 3.6 % to 1.1 % at 30 °C. For instance, the first resonance frequency in Fig.
7.4(b) has moved from 1.223 MHz to 1.305 MHz in Fig. 7.9. Such a shift can be interpreted
through total capacitance changes of the test object due to the dry-out process which will
be discussed in the last Section of this Chapter.
Further examination of the FRA traces in Fig. 7.9 and Fig. 7.10 reveal that the minimum
and maximum resonance frequencies have moved to lower frequencies as the
temperature increases. This movement is less than the movement of the ‘wet’ test object
spectra for 30 and 90 °C. Tables 7.6 and 7.7 provide detailed information of deviated
frequencies for FRA traces at 30 and 90 °C and also total capacitance variation for some of
the identical frequencies of HV and LV winding traces, respectively.
Based on the values reported in Tables 7.6 and 7.7, average capacitance deviations for HV
and LV windings were calculated as 2.37 % and 2.0 %, respectively. In addition, Fig. 7.11
shows that as the temperature increases the moisture content of the paper insulation is
changed from 1.1 % at 30 °C toward 0.4 % at 90 °C. Also, the oil moisture content has
increased from 3 to 21 ppm at 30 and 90 °C, respectively.
Technically speaking, from 30 to 90 °C the reduction of 0.7 % moisture content in paper
insulation caused around 2.18 % deviation in the total capacitance value for the ‘dry’ test
object, while these values were 1.9 % (WCP) and 6.22 % for the ‘wet’ test object.
Based on the result achieved from study of the test object under different moisture
contents, it is possible to calculate the total capacitance variation through FRA spectrum
deviation for 0.5 % moisture diffusion in paper insulation using linear interpolation (see
Table 7.8).
Table 7.8. FRA deviation and total capacitance variation for 0.5 % WCP change.
Model Transformer WCP Change FRA Spectrum Deviation (FSD) ΔC%

Wet 0.5 % 0.81 % 1.63 %
Dry 0.5 % 0.77 % 1.55 %
According to Table 7.8, each 0.5 % moisture migration from the paper insulation can cause
0.79 % displacement in FRA spectrum (as an average value for FSD) towards higher or
lower frequencies. In fact, 0.5 % moisture migration from the paper into the oil insulation
will lead to FRA spectrum moving to lower frequencies, while the same value of moisture
migration from the oil to paper insulation results in 0.79 % FRA movement to higher
frequencies.
152
DDF for wet and dry test objects as a factor showing the insulation quality was measured
and compared to the other results (see Table 7.9). When examining this table, it is obvious
that the DDF values for wet and dry test objects are completely different. In addition, as
the DDF increases the FRA trace is shifting to lower frequencies, and vice versa.
Table 7.9. Measured DDF at 5 kV before and after dry-out process.
Temperature 30 °C 40 °C 50 °C 60 °C 70 °C 80 °C 90 °C
‘Wet’ test
0.0203 0.0319 0.0484 0.0721 0.1119 0.1858 0.2849
object
‘Dry’ test
0.0028 0.0029 0.0044 0.0069 0.0088 0.0147 0.0210
object
7.3.6 Case Study 3 (Three-Phase Transformer)
To examine the above-mentioned result, a three-phase two-winding core type 20/0.4 kV,
1.6 MVA transformer was taken as another test object. The windings of this transformer
were homogenous conventional disk type. It was a spare transformer which was never in
service and had been kept in stock as reserve for a long time. At first, the oil sample was
taken from the transformer sampling valve and WCO and WCP were derived using KFT
method and MIT equilibrium curves, respectively (the oil temperature was 10 °C). These
values were 3 ppm and 3 %, respectively. Then, frequency response traces for HV
windings were recorded. To heat up the transformer in order to study the temperature
and moisture impacts, the secondary side of the test object was short-circuited and the
voltage was increased through the primary side to reach the nominal current on the
secondary side. It took 8 hours to reach to 60 °C for the test object. After that, the test
object was left for 48 hours under this condition to reach moisture equilibrium between
oil and paper insulations. The oil sample was then taken and FRA spectra were re-
recorded. WCO achieved was 26 ppm and WCP was derived as 1.9 %. Figure 7.12 shows
the FRA spectra for phase U on HV side at 10 and 60 °C. In addition, Table 7.10 provides
detailed information of the deviated frequencies in Fig. 7.12.
Based on the results, the total moisture variation for the paper insulation of the
transformer is calculated as 1.1 % from 10 to 60 °C. This variation has caused almost 1.5
% alteration in FRA trace and 3.22 % capacitance changes. Based on this ratio, it seems not
far from the fact that 1.49 % deviation in total capacitance is estimated for 0.5 % moisture
migration from the paper insulation in this transformer. This in turn almost verifies the
data achieved in Table 7.8.
153
0
-64 53.67 kHz at (at 10 °C)
-66 52.83 kHz at (at 60 °C)
-10-68
-70
Magnitude [dB] -72 37.66 kHz at (at 10 °C)

-20-74 37.09 kHz at (at 60 °C)
-76 16.22 kHz at (at 10 °C)
-78
15.98 kHz at (at 60 °C)
-30-80
Magnitude [dB]
-82
-84
-40 10
3 47.55 kHz at (at 10 °C)
Frequency [Hz]
46.78 kHz at (at 60 °C)
-50
-60
-70
-80 642.80 Hz at (at 10 °C) HV winding spectrum (at 60 °C)
635.10 Hz at (at 60 °C)
HV winding spectrum (at 10 °C)
-90 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 7.12. HV winding spectra at 10 and 60 °C (1.6 MVA transformer).
Table 7.10. HV winding capacitance ratio, anti-resonance and resonance frequencies for quoted points in
Fig.7.12.
Frequency f1(30°C) f2(90°C) f1/f2 C2/C1=( f1/f2 )2 ΔC%

1st minimum 642.80 Hz 635.10 Hz 1.0121 1.0243 2.4 %
1st maximum 37.66 kHz 37.09 kHz 1.0153 1.0427 4.2 %
2nd minimum 47.55 kHz 46.78 kHz 1.0164 1.0330 3.3 %
3rd maximum 16.22 kHz 15.98 kHz 1.0150 1.0302 3.0 %
2nd maximum 53.67 kHz 52.83 kHz 1.0159 1.0320 3.2 %
7.4 Verification of Practical Results Using Modelling and Simulation
A mathematical model using travelling wave theory had been developed in Chapter 4. This
model was verified through practical measurement in the same Chapter, and the transfer
function was given by (4.16).
This model is utilized in the current Chapter to examine the main reason of the FRA
spectrum deviation due to changes in the insulation characteristics.
On one hand, to verify the practical measurement through the theoretical approach and
estimate the main reason for a change in FRA response due to moisture and temperature
variations, each and every fundamental quantity controlling the FRA spectrum should be
studied. On the other hand, how to distinguish the winding inductance and capacitance and
resistance from one impedance measurement at specific frequency and ultimately through
the transfer function in frequency domain would be a challenge.
154
In fact, changes of FRA spectrum due to temperature and moisture variations are influenced
through changes of the inductance, capacitance and resistance in the winding impedance. In
this work, a simulation on frequency response of the model winding showed that the
winding resistance variation will lead to insignificant deviation in very low and high
frequencies in the FRA spectrum, while the resonant peak as well as spectral shape
remained unchanged (see Fig. 6.18). This work was verified through a practical study on
changes of FRA setup connections. Indeed, a bad, loose or oxidized connection in clamping
leads of the FRA test setup will change the resistivity of the setup, and this could be a model
of winding resistance variation for the winding under test. Frequency response
measurement on this case indicated similar results as to what happened in simulation study.
Very low and high frequencies were changed, while the resonant peaks remained
unchanged (see Fig. 6.19).
In addition, the results provided in Tables 7.2 and 7.3 on resistance variation showed that
the winding resistance alteration in various temperatures is insignificant in relation to the
impedance (combined with reactance). Hence, this hypothesis that FRA spectra are
influenced by winding resistance lost ground. Therefore, the remained reactance within
measured impedance in frequency response has almost certainly affected through the
inductance and capacitance.
The inductance alteration through the temperature variation was insignificant in Tables 7.2
and 7.3. On the other hand, it was shown that the low-frequency band of FRA spectrum is
influenced by the winding inductance (self- and mutual-inductance). Careful examination of
Figures 7.4(a) and 7.5(a) reveals that despite temperature and moisture variation, the low-
frequency band of spectra which is quite relevant to inductance has not deviated from the
origin. Also, shifting of local resonances has occurred in the mid- and high-frequency bands
in both graphs. This suggests that inductance deviation in this circumstance is negligible.
Investigation on resistance and inductance suggest that FRA spectrum deviation due to the
temperature and moisture variation can be initiated by capacitance variation.
In fact, in a capacitance model of insulation system, the current densities of bounded ions Jε
and free ions Jσ traversing a medium are obtained as:
J  D  j 0 r E , J   E (7.8)
t
where is the electric displacement, εr denotes relative permittivity of dielectric, ε0 is the
absolute permittivity of vacuum, is the electric field and σ designates dielectric
conductivity. The total current density traversing medium is then given by:
155
 E  ( j 0 r   )E  j 0 ( r  j  )E (7.9)

 0
χ is the electric susceptibility. Then, the complex permittivity is obtained as:
 *     r  j    r  j   C() (7.10)
0  0 C0
and current traversing the medium is given by:
I  jC0 (   j )U
insul insul (7.11)
where Uinsul is the voltage across the insulator, εʹ is the real part of the complex permittivity
related to the stored energy in the medium, εʺ is the imaginary part of the complex
permittivity related to the dissipation (or loss) of energy within the medium, C(ω) is the
complex capacitance with the dielectric present, and C0 denotes the capacitance without the
dielectric.
Therefore capacitance behaviour comes directly through dielectric behaviour, and

specifically dependent on dielectric permittivity and conductivity. Any change in these
factors will be reflected as a kind of change in capacitance. In addition, dielectric behaviour
usually depends on the variation of εʹ and εʺ with frequency, composition, temperature, and
voltage. In this study, the voltage remained unchanged for entire experiments; thus, just the
changes in frequency, temperature and composition need to be taken into consideration.
In the case of frequency and its variation, the real part of the dielectric permittivity is given
by ‎[116]:
 0   
     (7.12)
1   2 2
and the imaginary part (loss factor) is given by ‎[116]:
(     )
   0
(7.13)
1   2 2
where ε∞ and έ0 are the infinite-frequency dielectric constant and static dielectric constant
respectively, and τ is the dielectric time constant.
The real part of complex permittivity of dielectric to which (7.12) applies should always
decrease with frequency to reach ε∞, while εʺ tends to zero when ω approaches infinity. In
156
fact, εʹ approaches limiting values and εʺ experiences maximum value at ωτ=1 and negligible
values at ω=0 or ω=∞.
Hence, effective frequency band on εʺ is completely related to the dielectric time constant τ.
For an equal thickness of oil-paper composite insulations in transformer, τ is given by
(7.14):
 paper   oil
 (7.14)
 paper   oil
where, εpaper and εoil are the paper and oil insulation permittivity, and σpaper and σoil are the
paper and oil insulation conductivity, respectively.
For study at high frequencies (1 MHz- 20 MHz) similar to what happened in this study and
shown in Figures 7.4(a) and 7.5(a); the changes in the direction of the applied field are so
fast that the ions do not have to time to move an appreciable distance from their equilibrium
position before the direction of applied field is reversed ‎[116]. This could significantly
influence the contribution of ions (τ) and subsequently εʺ at high frequencies in the
dielectric structure. Thus, from frequency point of view, the impact of εʺ seems to be
insignificant in practice and εʹ tends to a determined value (ε∞).
Apart from frequency variation, temperature has its own impact on the real and imaginary
parts of the complex permittivity, whereas moisture content variation could be classified
into the composition effect.
In this regard, the study by Abeywichrama et al ‎[16] on transformer pressboard shows that
at a specific frequency if the temperature increases, then εʹ and εʺ will be increased.
Furthermore, if the moisture content decreases, then εʹ and εʺ will be decreased.
Another study by the same authors ‎[117] explains that the moisture variation mostly
influences the real part of permittivity εʹ, and the temperature affects mostly imaginary part
εʺ. In ‎[117], it is also discussed that the changes of εʺ due to the moisture variation is
considerable for the frequency below 100 Hz, while this is insignificant at frequencies higher
than 10 kHz. In addition as far as moisture increases, εʹ will increase significantly below 1
Hz, whereas weak increase can be observed above 100 Hz for this parameter ‎[117]. A
careful study on moisture variation in pressboard has been also reported in ‎[118] from an
original work by ‎[119]. It has been shown that moisture increment in the pressboard will
increase εʹ and εʺ specifically at low frequencies.
It is worth noting that the entire frequency band for study conducted in ‎[16] and ‎[117] is
100 Hz to 1 MHz, and in ‎[118] is 0.1 mHz to 1 kHz.
157
In our study, since temperature is changed from a lower value (30 °C) to reach a higher
value (90 °C), the moisture migrates from the paper into the oil insulation. In fact, when the
water in paper insulation decreases, the oil water content increases and it is not practically
possible to separate these reactions for interpretation. Furthermore, the separation of εʹ and
εʺ would be another challenge. Therefore, the introduced model in ‎[118] for composite
dielectric permittivity in transformer is used to address the issue.
In transformer design, the major part of the oil/paper insulation system is mostly
concentrated between the concentric HV and LV windings. Hence, significant amount of
water is available in this area. To study the transformer water content, this area is crucial
and should be investigated precisely.
Paper b/2
Cpaper
Oil Spacer
1-b
Coil Cspacer
Paper Cpaper b/2
1-a a
Figure 7.13. Schematic of the paper content and the spacer coverage in the insulation duct, taken and modified
[118].
In [118], it has been stated that for modelling purposes of this part it is sufficient to
represent the insulation structure by the relative amount of spacers and barriers in the duct
as it is depicted in Fig. 7.13. The major water content is absorbed by insulation, namely
pressboard barriers. If the thickness of pressboard barriers is thin enough to be close to
paper insulation thickness; then, it could be argued that instead of pressboard barriers the
paper permittivity plays major role in the model (see Fig. 7.13). Otherwise, paper
permittivity should be replaced by barrier permittivity in (7.15) and Cpaper should be
changed by Cbarrier in Fig. 7.13. With this assumption, the composite dielectric permittivity is
given by [118]:
 (,T )  a  1 a
1 b  b 1 b  b (7.15)
 spacer  paper  oil  paper
where εspacer denotes spacer permittivity and it is slightly higher than paper permittivity
εpaper, and a and b are shown in Fig. 7.13.
158
Since accurate measurement of the real and imaginary parts of the complex permittivity at
very high frequencies (f > 1 MHz) is not available, it has been assumed that εʹ and εʺ
behave as in low frequencies (f < 1 MHz). Knowing that when temperature and moisture
are changed, deviation of εʹ and εʺ from the origin under such experiment (f > 1 MHz) are
considerably less than that experienced at low frequencies (f < 1 MHz).
Based on this, if assuming the temperature of the test object increases, the moisture content
in paper insulation decreases and the oil water content increases accordingly. Hence, εoil will
take a greater value and εpaper and εspacer will experience less values. On the other hand εoil <
εpaper; therefore under such circumstance, ε in (7.15) becomes larger. This increment could
be even intensified when the oil insulation thickness is greater than paper insulation within
the composite insulation medium.
All in all, increment of ε could almost certainly influence the capacitance and conductance.
Therefore, the conductance and capacitance were changed in the model and the frequency
responses were simulated.
In the case of conductance variation, simulation results showed that the conductance
increment for both turn-to-turn and shunt-to-ground conductance causes slight damping in
resonance peak magnitudes, while resonant frequencies and FRA spectrum trend remained
unchanged. This impact was more significant in shunt conductance variation (G) rather than
turn-to-turn conductance (g).
In the case of capacitance variation, at first the reference FRA spectrum was simulated for
the HV winding. Then the relative permittivity in shunt and series capacitances was changed
gradually and different spectra were simulated and compared to the reference spectrum.
Simulation results showed that a slight change (around 6.2 %) in series capacitance does not
cause considerable deviation in FRA spectrum, while significant change (around 200 %) will
slightly move the FRA trace in the mid and high frequencies. This effect became more
significant when the shunt capacitance was changed. The increment of just 6.2 % of the
relative permittivity of the shunt capacitance caused the FRA resonant peaks moved to
lower frequencies for 3 % (see Fig. 7.14).
These results in turn suggest that FRA spectrum deviation due to the temperature and
moisture variation comes primarily through the shunt capacitance rather than series
capacitance. Therefore, the right hand side columns in Tables 7.4, 7.5, 7.6 and 7.7 are
focused on the shunt capacitance variation rather than series capacitance. More precise
study on this case could be conducted through εʹ and εʺ measurement at high frequencies (f
> 1 MHz).
159
0
Cg 3 % deviation in resonant peak
-10 1.062 x Cg 4.376 MHz 4.510 MHz
-20
-30
Magnitude [dB]
-40
-50
-60
-70
-80
-90
6 7
10 10
Frequency [Hz]
Figure 7.14. Simulated FRA spectra for model transformer (shunt capacitance deviation).
7.5 Influence of Temperature and Moisture Content on FRA Statistical

Indicators
It was discussed in Chapter 2 that a common method of interpreting FRA data is to use
statistical indicators (indices), particularly the correlation coefficient (CC) and standard
deviation (SD). In order to evaluate the capability of these indicators in FRA spectrum
interpretation due to temperature and moisture variation, all measured results in the
above case studies were examined by CC and SD. Calculated values of CC and SD for the
FRA spectra of the model transformer under different moisture contents (‘wet’ and ‘dry’)
at 30 and 90 °C as well as the power-rated transformer at 10 and 60 °C are given in Table
7.11. It can be seen that CC and SD indicate deformation for the model transformer (for
both ‘wet’ and ‘dry’ cases) but in reality, its windings have not deformed (see Table 2.6 for
criteria). On the other hand, results for the power-rated transformer indicate that its
condition is ‘normal’.
Table 7.11. Calculated statistical indices.
Test Object CC SD
‘Wet’ Model Transformer, LV Spectra (at 30 °C and 90 °C) 0.9967 1.4989
‘Wet’ Model Transformer, HV Spectra (at 30 °C and 90 °C) 0.9974 1.5258
‘Dry’ Model Transformer, LV Spectra (at 30 °C and 90 °C) 0.9967 1.4989
‘Dry’ Model Transformer, HV Spectra (at 30 °C and 90 °C) 0.9991 0.8841
1.6 MVA Real Transformer, HV Spectra (at 10 °C and 60 °C) 0.9999 0.6000
160
This study reveals that statistical indices (CC and SD) do not have enough accuracy in FRA
spectrum interpretation when FRA traces have been taken at different temperature and
moisture contents. Hence, the existing available statistical indices need to be modified.
7.6 Practical Solution to Modify Statistical Indicators
According to results in the last Section, to obtain the maximum accuracy in transformer
frequency response evaluation, the statistical indicators need to be modified and the
moisture and temperature variations are brought into consideration. In fact how to
distinguish the temperature and humidity impacts on the FRA spectrum from the winding
deformation is the challenge.
A study by Ryder in ‎[28] did not focus on the moisture migration and its impact on the
FRA spectrum; however, as a caution he advised using inter-phase comparisons for those
cases suspected of experiencing relative humidity changes. It means that to recognize the
relative humidity changes in transformer FRA signature, the FRA spectra recorded for
different phases on HV side can be compared together and their discrepancy is taken as
the influence of humidity changes. Similar procedure can be performed for LV side.
However, it was emphasized that this suggestion is feasible only for three-phase
transformers ‎[28].
The present study believes the recommended solution by ‎[28] has fundamental
limitations. Even for the lateral windings in a three-phase transformer (phases A and C),
having similar frequency responses is often not achievable in practice. There are many
cases in which lateral FRA fingerprints are quite dissimilar to compare. Also, the
assessment of the middle winding frequency response will remain a challenge. Its
spectrum is not comparable to the others.
Ryder’s suggestion could be extended into inter-winding instead of inter-phase

comparison. Inter-winding comparison can be performed between the HV and LV
windings for each and every phase independently. In fact, a differential spectrum (Xi(HV)-
Xi(LV)) can be calculated using HV and LV winding frequency response spectra and called
DFRA. DFRA spectrum which comes through the difference between HV and LV winding
spectra should remain unchanged and also not altered due to moisture and temperature
changes. Any changes in paper moisture or temperature in HV winding would occur for
the LV winding as well. As far as DFRA spectrum is not changed for the transformer entire
life, it means that transformer humidity has remained unchanged and the measured FRA
spectra can be examined through statistical indicators to explore winding deformation.
161
This approach seems to be more feasible than what has been recommended in ‎[28], but it
has still a drawback. The volume of crepe paper insulation for HV and LV conductors as
well as paper insulation thickness for each and every winding might be different. In
addition, a transformer might have an enamelled winding in LV side and a paper insulated
winding in HV side. Thus the water absorbed through the HV and LV windings in a phase
would be different and result in an incorrect prognosis. Therefore, this approach also
appears to be not quite accurate. Hence, a comprehensive practical solution which
provides an independent result for each and every winding would be preferable. This
solution is discussed hereinafter.
Indeed before the evaluation of FRA spectra through the statistical indicators, a
preliminary calculation should be performed to distinguish insulation characteristic
influence from winding deformation. Figure 7.15 presents the flowchart of a new
technique to address this issue. This procedure distinguishes the insulation impact from
the mechanical impact on the frequency response of transformer winding. The reference
and measured spectra are numerically processed to determine whether further evaluation
by statistical indicators can proceed, or other actions are required (including possible
outcome of ‘no action’).
In this chart, α and β are considered as the lower and upper limits of FRA trace deviation
and will be determined only through the FRA signature (reference trace). At first, the
moisture content of the paper insulation should be measured during the FRA signature
measurement. Next, the measured value for the paper moisture is taken as the reference
value (WCPref) in percent for FRA trace signature. Afterwards, deviation in the measured
FRA spectrum in Bode diagram due to the moisture migration from paper insulation is
calculated (FSD), and the moisture diffusion (WCP change) is derived (see Table 7.8).
Then, α and β can be determined through (7.16) and (7.17) considering 0.5 % (Wll) and 4
% (Wul) as the lower and upper criteria for moisture content of the transformer paper
insulation (different standards or guidelines may recommend different criteria for
maximum and minimum moisture content in transformer paper insulation).
Comparing Rn ≜ R{Xi, Yi}n in Fig. 7.15, α and β, appropriate decision can then be made. In
fact, Rn should remain between α and β. In order to clarify the procedure, Appendix G
provides practical examples on the case studied in Chapter 7.
  1  FSD
Wll  WCPref  (7.16)
WCP Change  100
162
  1  FSD
Wul  WCPref  (7.17)
WCP Change  100
where Wll and Wul are the lower and upper limits of the paper moisture content in percent,
WCPref is the reference paper humidity in percent and should be measured during FRA
fingerprint measurement, FSD is the average FRA Spectrum Deviation for 0.5% moisture
variation in paper insulation, WCP Change is the per unit water content change in paper
insulation and equal to 0.5 in Table 7.8.
163
Start
Reference Measured
Spectrum Spectrum
Xi Yi
Derivative Curve Calculation

X i Y
Xˆ i  , Yî  i
fi fi
Zero Crossing Recognition

S{Xˆ i }  Sign( Xˆ i ) , S{Yî }  Sign(Yî )
Z{Xˆ i }n  S{Xˆ i 1}  S{Xˆ i }  0 , Z{Yˆ }  S{Yˆ }  S{Yˆ }  0
i m i 1 i
No FRA Statistical
n=m
Indicators
Yes
Ratio Calculation
Z {Yî }m
R{ Xˆ i , Yî }n 
Z { Xˆ i }n
No Trip, Humidity
  R{ Xˆ i ,Yî }n   Investigation is
Required
Yes
No Action is
Required
End
Figure 7.15. The chart on preliminary calculation on FRA traces to distinguish insulation deviation from
winding deformation; Xi and Yi were defined in Chapter 2.
164
7.7 Transformer Winding Dry-out Influence on Frequency Response Trace
Figure 7.16. Glassy model transformer.
The achievement in the last Sections revealed that if the paper water content is changed,
then the frequency response of transformer winding will shift horizontally towards lower
or higher frequencies. The results also shown that each 0.5 % moisture migration from the
paper insulation will lead to almost 0.79 % deviation in FRA trace. This deviation was
observable and calculable for resonant frequencies. In practice, as a common transformer
active part treatment, the transformer dry-out process removes significant amount of
moisture from the paper insulation and causes the paper moisture content to change.
Therefore, it could have significant impact on FRA trace due to paper moisture diffusion.
In order to address this hypothesis, the frequency response traces of wet and dry model
transformer are shown in Fig. 7.17 for HV and LV windings, respectively.
According to Fig. 7.17(a) and Fig. 7.17(b), moving from low frequencies to 800 kHz all of
the traces are perfectly matched. Moving from 800 kHz to the higher frequencies, the
discrepancy between the traces becomes obvious. Since, the test object moisture content
has changed from 4 % to 1 %, the FRA traces have slightly moved to higher frequencies for
both LV and HV windings.
Total capacitance variation of transformer windings due to the moisture diffusion could be
considered as the major factor influenced FRA trace before and after the dry-out process.
To calculate the total capacitance variation, some of the resonances and anti-resonances
have been highlighted in Figures 7.17(a) and 7.17(b).
Tables 7.12 and 7.13 show the numerical values of deviated frequencies for 4 % and 1 %
water content of the paper (WCP), where f1 and f2 represent the frequencies of anti-
resonances (minima) and resonances (maxima) at 4 % and 1% WCP, respectively.
165
0
4.651 MHz, WCP 1%, at 23 °C
-10 4.313 MHz, WCP 4%, at 23 °C
1.899 MHz, WCP 1%, at 23 °C

-20
1.780 MHz, WCP %, at 23 °C
-30
Magnitude [dB]
-40
-50
2.006 MHz, WCP 1%, at 23 °C
-60 1.861 MHz, WCP 4%, at 23 °C
-70 1.222 MHz, WCP 4%, at 23 °C
-80
HV winding spectrum (at 23 °C) - After dry-out, WCP 1% 1.318 MHz, WCP 1%, at 23 °C
HV winding spectrum (at 23 °C) - Before dry-out, WCP 4%
-90 4 5 6 7
10 10 10 10
Frequency [Hz]
(a)
0 7.483 MHz, WCP 1%

at 23 ºC
6.942 MHz, WCP 4%
at 23 ºC
-10 1.119 MHz, WCP 4%

at 23 ºC 1.206 MHz, WCP 1%
at 23 ºC
-20
Magnitude [dB]
-30
-40
0.947 MHz, WCP 4%

at 23 ºC 2.588 MHz, WCP 1%
-50 at 23 ºC
LV winding spectrum (at 23 °C) - Before dry-out, WCP 4% 2.427 MHz, WCP 4%
1.022 MHz, WCP 1%
at 23 ºC
LV winding spectrum (at 23 °C) - After dry-out, WCP 1% at 23 ºC
-60 4 5 6 7
10 10 10 10
Frequency [Hz]
(b)
Figure 7.17. FRA traces for glassy test object (a) HV winding spectra before and after dry-out (frequency
band 3 kHz-20 MHz, 0 dB < 3 kHz), (b) LV winding spectra before and after dry-out (frequency band 6 kHz-
20 MHz, 0 dB < 6 kHz). The measurements have been performed for oil-filled model transformer.
166
Table 7.12. Capacitance ratio, anti-resonance and resonance frequencies for quoted points in Fig. 7.16(a).
Frequency f1 (WCP 4%) f2 (WCP 1%) C2/C1=( f1/f2 )2

1st minimum 1.222 MHz 1.318 MHz 0.86
1st maximum 1.780 MHz 1.899 MHz 0.88
2nd minimum 1.861 MHz 2.006 MHz 0.86
2nd maximum 4.313 MHz 4.651 MHz 0.86
Table 7.13. Capacitance ratio, anti-resonance and resonance frequencies for quoted points in Fig. 7.16(b).
Frequency f1 (WCP 4%) f2 (WCP 1%) C2/C1=( f1/f2 )2

1st minimum 0.947 MHz 1.022 MHz 0.86
1st maximum 1.119 MHz 1.206 MHz 0.86
2nd minimum 2.427 MHz 2.588 MHz 0.88
2nd maximum 6.942 MHz 7.483 MHz 0.86
Calculation of capacitance ratios at 4 % and 1 % WCP show that due to the moisture
diffusion, the total capacitances of the windings have changed around 14%. These changes
in the winding capacitance can be interpreted through 3 % moisture diffusion from the
paper insulation due to the dry-out process.
7.8 Conclusion
The influence of temperature and moisture content on the frequency response trace of
transformer winding was discussed in detail. This study indicated that transformer
temperature and moisture variation can lead to FRA spectrum deviation. In fact, moisture
migration from the paper into the oil insulation will cause the FRA spectrum shifting
horizontally (frequency-axis) to lower frequencies, whilst moisture diffusion from the oil
into the paper insulation will shift the FRA spectrum to higher frequencies. Detailed study
on moisture migration for different test objects showed that changes of 0.5 % moisture in
paper insulation will move FRA spectrum for 0.79 %. Simulation results indicated that this
movement is significantly affected by changes of transformer winding shunt capacitance.
In the case of transformer dry-out evaluation, the total capacitance deviation from the
reference value for wet and dry paper was estimated. In summary, the conclusion is that
the FRA appears to be able to provide significant information on moisture migration. It
may even be an effective tool to double–check the efficiency of transformer dry-out
process. Thus, an FRA measurement could be performed before and after a transformer
dry-out process to ensure moisture diffusion. However, it is emphasized again that more
work is necessary for establishing a reliable moisture migration recognition method
167
through FRA. Therefore, it is recommended that FRA measurement be performed on

transformers (along with other methods) before and after the dry-out process as a part of
moisture diffusion assessment.
Study on the statistical indices revealed that these indicators should be modified or their
criteria must be revised to distinguish winding deformation influence from insulation
characteristic impacts on FRA trace. A method was recommended for the first time to
address this issue.
Apart from recommended method, a possible solution to distinguish winding deformation
from insulation characteristic effects on FRA spectrum is implementing on-line FRA
measurement. In fact, FRA spectrum will change gradually as the insulation characteristic
is changed. On the other hand, winding deformation will trigger a sudden shift of the FRA
graph. On-line FRA measurement monitors the winding frequency response continuously,
and thus is able to distinguish the gradual change of FRA spectrum due to the insulation
characteristic variation from the fast displacement of FRA trace due to the winding
deformation. However, more work is necessary to establish a comprehensive technique.
To open the discussion for future study, the online FRA and its setup is discussed in the
next Chapter in detail.
168
Chapter 8. On-line Transformer Winding Deformation Diagnosis
Chapter 8
On-line Transformer Winding

Deformation Diagnosis
8.1 Introduction
In Chapter 7, an off-line technique was introduced to distinguish the impacts on FRA

traces caused by insulation characteristic changes as against winding deformation. It was
also discussed and explained that on-line FRA application is another possible solution for
this discrimination.
In fact, on-line FRA application not only helps to address this issue but it is also expected
to be a logical evolution in transformer winding deformation diagnosis. In order to
implement on-line FRA successfully, all aspects of the monitoring system should be
regarded. Technical issues and practical challenges should be addressed and overcome.
To date, transformer tank vibration [120]-[123], communication technique using scatter

parameters [124]-[126], current deformation coefficient [127], ultrasonic method [128],
short circuit impedance [129]-[134] and winding stray reactance [135]-[137], on-line
Transfer Function (TF) using time domain or frequency domain measurements [10], [30],
[33]-[35], [138]-[139] have been introduced as advanced on-line methods for recognition
of transformer winding deformation or displacement. Although they have been all
practically investigated, there is no industrial evidence of permanent implementation.
This Chapter gives a brief review on recommended online methods, and then extensively
discusses on-line FRA application. On-line FRA setup is introduced and practical
challenges are highlighted and discussed in detail. How to achieve the maximum
information through on-line FRA setup is considered a fundamental challenge. This
challenge is studied practically through different test objects and a possible solution is
recommended.
169
8.2 Advanced Methods in On-line Transformer Winding Deformation

Diagnosis
8.2.1 Vibration Method
Transformer vibration can be considered to be repetitive movement of transformer inner

parts that are enclosed within the transformer tank. This movement occurs around a
reference position. The reference position is where the transformer attains once it is out of
service. Vibration might be interpreted by using parameters such as winding
displacement, velocity and acceleration.
Based on this, [120]-[123] have introduced an on-line method. These studies show that
the extent of transformer tank vibration depends on voltage square and current square.
Furthermore, the winding vibration frequency is 100 Hz when the fundamental power
frequency is 50 Hz. Core vibration which is caused by magneto striction and magnetic
forces is as (8.1) for the proposed model in [120], [122].
tank 100Hz    to  i100

2
 (  to )u100
2
(8.1)
υtank-100Hz is the frequency tank vibration at 100 Hz, i2100 is the current square harmonic at
100 Hz, u2100 is the voltage square harmonic at 100 Hz, and θto is the oil temperature
measured at the top of the tank. α, β, γ and δ are proposed coefficient in [120].
Considering (8.1) and alteration in υtank-100Hz, the transformer tank vibration has been
recommended to be considered as an on-line transformer winding deformation diagnosis
method.
8.2.2 Communication Method
The communication method which is introduced in the literature [124]-[126] is based on

scattering parameters. The magnitude and phase of scattering parameters for normal
transformer winding are measured by several antennas as finger print. Proposed antennas
could be placed outside or inside the transformer tank. In this method, the mean absolute
magnitude distance (MAMD) and mean absolute phase distance (MAPD) are introduced as
displacement indices. MAMD and MAPD have been calculated in [126] and are as (8.2) and
(8.3), respectively:
n
 || Si |  | Sref ||
(8.2)
MAMD  i 1
n
170
n
 | Si  Sref |
(8.3)
MAPD  i 1
n
where Si is the measured scatter parameters, Sref is the scatter parameters. As mentioned
in [124]-[126], any kind of transformer winding deformation can cause the above-
mentioned indices altered and hence deformation detected.
8.2.3 Current Deformation Coefficient Method
This method has been introduced by [127]. A high frequency low voltage signal is applied
to live power system line along with power frequency signal when the standard practices
of connection are followed. The line-end and neutral-end high frequency currents are
continuously measured using isolated precision current probes and digital filtering
technique [127]. Associated capacitive reactance is changed due to the transformer
winding deformation and this change is reflected in deviations of high frequency terminal
currents from the fingerprint. When these deviations are measured, the ratio of deviations
at the two ends is calculated. Hence, current deviation coefficient (CDC) is introduced and
given by [127]:
 I1H  I1H 
CDC  log10   (8.4)
 I 2 H  I 2 H


I1H and I2H are the fingerprint values of measured terminal currents at the selected high
frequencies, and I'1H and I'2H terminal current values at the winding terminals after
deformation.
8.2.4 Ultrasonic Method
Ultrasound is a sound with a frequency greater than the upper limit of human hearing, i.e.
~20 kHz. In this method introduced in [128], an ultrasonic signal is used as the reference
signal. The basis of this method concentrates on ultrasound reflection due to the non-
matching acoustic impedance between the oil and the winding.
8.2.5 Online Short Circuit Impedance and Winding Stray Reactance Method
Based on the literature [129]-[137], the measured short circuit impedance of a power
transformer can be compared to the value which has already been recorded. In fact, on-
line short circuit impedance method relies on time based comparison. Considering
171
equation (2.5), short circuit impedance of a transformer is related to the windings

configuration and distance between windings.
Off-line transformer short circuit impedance measurement is performed when the

secondary side of transformer is shorted and voltage excitation is carried out at the
primary side. Since secondary short circuit setup is not possible for energized
transformer, the following equation for on-line short circuit impedance calculation is
utilized. A two-port network as a model for transformer is considered as shown in Fig. 8.1.
I1 I2
Two-Port
U1 U2
Network
Figure 8.1. Two-port network.
Demonstrated parameters in Fig. 8.1 are defined as:
 U1   Z11 Z12  I1 
    (8.5)
U   Z Z22  
 2   21  I 2 
Z11 is the open circuit input impedance, Z12 is the open circuit reverse transfer impedance,
Z22 is the open circuit output impedance, and Z21 is the open circuit forward transfer
impedance.
When short circuit happens, U2=0 and (8.5) becomes:
U1  Z11I1  Z12 I2
(8.6)
0  Z21I1  Z22 I2
Hence, the short circuit impedance (Zsc) is given by:
Z Z
Z sc  Z11  12 21 (8.7)
Z22
Any deviation in Z11, Z22, Z12 and Z21 will cause Zsc to change. All impedances can be
calculated using (8.8) and (8.9):
 U1  Z11I1  Z12 I 2
 t1 t1 t1
 (8.8)
U 2  Z 21I1  Z 22 I 2
 t1 t1 t1
172
 U1  Z11I1  Z12 I 2
 t2 t2 t2
 (8.9)
U 2  Z21I1  Z22 I 2
 t2 t2 t2
U1t1 and U1t2 are the primary voltages in Fig. 8.1 corresponding to t1 and t2, respectively, I1t1
and I1t2 are the primary currents corresponding to t1 and t2, U2t1 and U2t2 are the secondary
voltages corresponding to t1 and t2, I2t1 and I2t2 are the secondary currents corresponding
to t1 and t2, and t represents the time of measurement.
Based on equations (8.8) and (8.9), proposed impedances can be calculated using equation
(8.10):
 Z11

Z12   U1t1 U1  I  I1  
 
t 2  2t 2 t2   1 
 .

  (8.10)
Z
 21 Z22  U 2 U2   I

I
I1   1 2 I I I
2t1 1t 2 
 t1 t 2  2t1 t1   t1 t 2 
When the left hand side matrix in (8.10) changes, Zsc will change. On-line short circuit
impedance measurement has been introduced as an advanced and economical method in
transformer winding deformation diagnosis. However, some researchers have argued on
the accuracy of this method.
8.2.6 On-line Frequency Response Analysis (On-line FRA)
While off-line FRA is now a common method for transformer winding deformation
diagnosis, its extension to on-line FRA measurement is still evolving. Since most power
transformers are equipped with capacitively-graded bushings, the transformer bushing
tap is suitable as an input point for low voltage signal injection during on-line FRA
measurement ‎[30].
Figure 8.2 shows the side cut-off of a capacitive bushing. The bushing tap is connected to
the last layer of capacitive grading which is brought out insulated at the bushing flange via
a small auxiliary bushing. This tap provides a much reduced terminal voltage due to the
capacitive divider. In fact, during transformer operation the transformer bushing tap is
grounded through the cover cap with internal spring. The cover has a contact socket into
which the contact pin locks when the cap is closed.
On-line FRA measurement requires opening the grounded bushing tap so to be able to
inject signal into the transformer without requiring a direct connection to the main feeder.
Hence, C1 and C2 values as a capacitive divider are important parameters during on-line
173
FRA measurement. Table 8.1, shows typical values of measured capacitors for 230 kV and
400 kV bushings.
Figure 8.2. Side cut-off of a capacitive bushing.
Table 8.1. Typical values for bushing capacitances [10].
Rating [kV] C1 [pF] C2 [pF] Ratio (C2/C1) Tap Voltage [kV]

230 608 6192 10.18 20.56
500 498 10021 20.12 23.67
Measured values for C1 and C2 by bushing manufacturers show that the division ratio
causes high voltage on the test tap if it is not grounded in operation time. Obviously, the
creepage distance and also air clearance between the bushing tap and the grounded flange
is not coordinated to withstand against estimated voltage as calculated by equation (8.11):
C1
Vtap  V (8.11)
C2  C1 phase
where Vtap is the bushing tap voltage, and Vphase denotes the phase voltage. In addition, the signal
generator device cannot tolerate high voltage due to its low insulation level. Therefore, it
is imperative to have an appropriate shunt impedance in parallel with C2 when the
bushing tap is not grounded during on-line measurement. This is illustrated in Fig. 8.3. The
divided voltage on the bushing tap is given by:
Z p || 1/ jC2 
Vtap  V (8.12)
 Z p ||
 1/ jC2  1/ jC1 phase
where Zp is the impedance in parallel with the bushing tap, and ω is angular frequency.
174
Furthermore, the capacitor values of the bushing remain relatively constant over a wide
frequency range. This should limit the adverse effect of the bushings’ own frequency
response masking the actual transformer signature [36]-[37]. This is discussed and
probable solution is recommended in the next subsection. Wye and delta circuits for on-
line frequency response measurement are shown in Fig. 8.4 and Fig. 8.5, respectively.
As it is demonstrated in Fig. 8.4, with wye connection, on-line FRA can be measured by
signal injection in the phase bushing tap and response can be recorded through the
neutral bushing tap. Also, it can be measured between two phases when the connection is
delta type as it is shown in Fig. 8.5.
8.3 Discussion
To reach to an acceptable level of technical satisfaction, there are number of gaps that
need to be addressed.
Obviously, each introduced on-line method has its own advantage and disadvantages and
what is considered as a challenge for a particular method arises from the disadvantages.
Figure 8.3. Paralleled impedance with bushing tap (test tap) on phase U.
175
ZN
Z1 Z2 Z3
U V W
C1N C1U C1V C1W
C2U C2V C2W

C2N
ZU ZV ZW
Figure 8.4. Wye circuits for on-line frequency response measurement.
Z1 Z2 Z3
U C1U V W
C1V C1W
C2U C2V C2W
ZU ZV ZW
Figure 8.5. Delta circuits for on-line frequency response measurement.
Many questions could be raised regarding the introduced on-line methods. As for on-line
transformer tank vibration measurement, external factors in practical environments will
affect the measurement considerably. Any kind of electro-dynamic forces present can
interfere with transformer vibration tests. In particular, most of transformer accessories
will lead to some vibration in transformer tank. Therefore, interpretation of transformer
tank vibration test result will be difficult.
As for the communication method in on-line transformer winding deformation diagnosis,
it is worthy to note that the transformer tank is grounded when the transformer is in
service. Hence, signal penetration through the transformer tank is not easily possible since
the whole of transformer active part is shielded by the tank. Even if the transmitter
sensors are placed inside the transformer tank, concentric windings will act as a shield for
each other. This method seems to be more suitable for reactors than transformers and
only if the sensors can be located inside the reactor tank. Still, result interpretation would
be a challenge due to complicated electromagnetic wave propagation.
Amongst all, short circuit impedance and frequency response measurement in frequency
domain have fewer drawbacks as compared to the other recommended methods. These
two methods are less affected by external factors, more economical and reliable to
176
perform on-line. They have also been industrially tested off-line and recognized as reliable
techniques, even though for on-line application, a number of challenges will be expected:
 The basic question is which method is more accurate? To verify the accuracy, SCI and
FRA should be applied on a faulty power transformer to identify the accuracy of each
method. In the absence of such a test, the more precise method when performed off-line
should be a logical choice for on-line application. This issue was studied and addressed in
Section 2.4 and it was concluded that FRA is superior to SCI.
 Some researchers believe that short circuit impedance is applicable for on-line
transformer deformation diagnosis because it can be readily obtained through voltage and
current measurement. However, others are concerned that the measured impedance has a
non-linear relationship with supply voltage as well as core magnetizing level. Therefore,
non-linear coefficient must be considered to achieve short circuit impedance [23]-[28]. In
addition, based on equation (2.5) it cannot recognize disk-to-disk axial movement as well
as tilting. .
On the other hand, in case of on-line FRA measurement; there are many complicated
practical challenges when this method is going to be performed on-line. In fact, on-line
FRA measurement concept is the same as off-line but problems arise as the transformer is
energized during measurement process:
 The crucial one is validity of on-line FRA measurement results. How to isolate the
response of external system from the winding frequency response is a challenge that must
be addressed.
 With off-line FRA measurements it has already been explored that transfer function of
the primary side of a transformer is different for open circuit and short circuit
configuration of the secondary in low frequencies [33]-[34]. With on-line FRA
measurements transformer is energized and secondary side is normally loaded. In this
case, load variation in operation time is inevitable. Variable load might be modeled as
parallel variable impedance with the secondary side of the transformer [31]. This in turn
will result in variations of frequency response measured data.
 On-line FRA measurement needs to be performed separately for each winding. In fact,
injected signals for three phase transformers must not be exercised simultaneously. This is
to avoid the problem of signal overlap and superposition. Using a multiplexer for signal
injection rather than individual signal generator for each phase might be an appropriate
practice.
 On-line frequency response measurement can be performed in time domain (LVI) or
frequency domain (FRA). On one hand time domain is much faster than frequency domain.
177
Impulse excitation, storage of samples, signal processing and calculation should take only
a few seconds [145], while duration of measurements in frequency domain usually is
about a few minutes. On the other hand, frequency domain result is more accurate than
that in time domain. Based on literatures, sometimes two different signals in time domain
may lead to the same frequency domain characteristic [146]. Therefore, frequency domain
measurement is more reliable. Depending on the environmental and operational
conditions and other factors involved choosing the appropriate method for every single
case would be a challenge.
 The effects of the input signal on the protection system while on-line FRA is conducted
should be explored in detail practically. As the voltage of the injected signal could be about
230 volts [30], its influence would be negligible.
 A number of studies have suggested that high frequency CT should be applied as an
output probe on neutral point (if exists) to measure the response. With this arrangement,
the CT frequency response would cause additional error in results and increase the
uncertainty of the main winding’s frequency response. Error and uncertainty calculation
formulas for winding and also CT are as (8.13) and (8.14):
er TF   er  winding   er (CT ) (8.13)
u(TF )  u(Winding )2  u(CT )2 (8.14)
where, er is the measurement error, and u denotes estimated uncertainty.

 As discussed, on-line frequency response measurement would be applicable for HV side
of transformer through the bushing tap (test tap). On the other hand, the low voltage side
of transformer often is oil bushing. Since the oil bushing does not have bushing tap, a
Rogowsky coil can be used as the sensor [35] and the output voltage is calculated as:
Nrog 0 A dI
rog
V  (8.15)
O rog dt
Nrog is the number of coil turns for Rogowsky coil, λrog is the length of the winding, and Arog
is the cross-section area of each small loop in Rogowsky coil.
This formula assumes the turns are evenly spaced and that these turns are small relative
to the radius of the coil itself. In addition, since a Rogowsky coil has an air core rather than
an iron core, it has a low inductance and can respond to fast-changing currents. Also,
because it has no iron core to saturate, it is highly linear even when subjected to large
currents. This coil would be appropriate regarding immunity against electromagnetic
178
interference because of uniformly spaced winding. However, its use will result in the
frequency response of this coil interfering with the transformer winding transfer function.
This interference should be minimized as much as possible.
 Since HV and LV transformer windings’ frequency responses are affected by self and
mutual inductances as well as series and shunt capacitances values, any changes in these
elements will lead to frequency response alteration, accordingly. Indeed, off-line
frequency response is mostly affected by self-inductance, series and shunt capacitances. In
fact, as in off-line measurement the test voltage is often applied to the primary winding
whilst the secondary winding is open circuit and without carrying any alternating current,
the mutual inductances between the windings do not contribute to frequency response
formation. The transformer winding mutual inductance is given by [92]:
M  q L1L2 (8.16)
where, L1 and L2 represent the self-inductance of primary and secondary windings; q is

coupling coefficient and depends on closeness of coupling between the windings.
On the other hand, in on-line transformer frequency response measurement, as both

windings are carrying current, the mutual inductances between the windings could have
significant impact on the frequency response. Therefore, in the event of a deformation in
LV (or HV) winding or the channel width between HV and LV windings, physical changes
will be reflected in HV (or LV) winding transfer function. However, if tilting or bending
occurs on LV (or HV) winding, the frequency response of HV (or LV) winding probably will
not change. Any deformation or displacement that alters the volt per turn value of the
transformer will change transformer core flux. In turn, the self and mutual inductances
will be affected specifically in low frequencies. This will lead to frequency response
variation. In fact, while the transformer is energized, its active parts consisting of the core,
HV and LV windings as well as radial and vertical channels in between can be considered
as one single integrated system.
 The last and probably the most important concern is how to get the maximum
information through online FRA test setup. This is as a significant challenge. The
remaining Sections in this Chapter specifically highlight and discuss this issue.
8.4 Problem Statement on Online FRA Setup
Based on the literature [20], the off-line transformer frequency response finger print is
commonly reported between 20 Hz and 2 MHz. This frequency band is capable of
179
providing significant information on transformer core and windings as well as clamping

structures. In fact as discussed earlier, each sub-band has its own meaning. Thus, having
entire FRA spectrum is quite important for its interpretation.
The study conducted in this thesis shows that using the on-line FRA measurement setup
will mask out the low-frequency data in FRA spectrum. In fact, the data obtained from the
off-line FRA measurement will not be fully retrieved through the on-line setup. The low-
frequency band of the on-line FRA spectrum shows very rapid fluctuations. It also exhibits
a rising trend instead of falling trend which is quite different to what expected for a
transformer. This issue is also discussed in a work by Behjat et al [32] through a study on a
35 kV power transformer. In [32], it raised the concern that “the major limitation of the
online method comes from the noise and environmental effects at lower frequencies”.
Moreover, it warned that “owing to noise and environmental effects, online measurement
of the winding transfer function was nearly impossible in the frequency range of lower
than 1 kHz”.
The current study believes that these fluctuations are not apparently due to the noise and
could be overcome through appropriate design of on-line circuit components. Therefore,
there should be an investigation to address this issue and obtain a proper on-line FRA test
setup having minimum influence on on-line FRA results.
8.5 Challenges with On-line FRA Setup
Based on the recommended on-line FRA measurement setup, the fundamental issue is the
bushing impact on the frequency response spectrum of transformer. In fact, when FRA
signal is going to be injected through the transformer bushing tap, the bushing capacitance
will certainly affect the FRA data.
If assuming the winding’s behaviour is quite inductive in the low frequencies; then, in on-
line FRA setup the bushing capacitance has series connection with the winding inductance.
This in turn causes the total circuit impedance to change, and consequently the FRA trend
altered, specifically, in the low-frequency region.
In the case of an oil bushing, a Rogowsky coil or a coupling capacitor is required. The idea
of coupling capacitor implementation instead of using the bushing tap comes from the
literatures ‎[30], ‎[34] and ‎[35].
Using coupling capacitor in on-line FRA setup will affect the transformer frequency
response. In this case, the on-line frequency response interpretation would be even more
complicated. This effect should be eliminated or minimized as much as possible. To
180
address this issue, recommended solutions to gain maximum information through on-line
FRA setup are discussed through the practical case studies, hereinafter. In this regard, the
first case study tries to highlight the issue for better understanding of the impact of
bushing capacitance on on-line FRA setup, the second and third case studies recommend a
solution, and the last two investigate its performance when applied to different test
objects.
8.6 Case Studies
In order to evaluate the on-line FRA measurement setup, two 66 kV, 25 MVA transformer
interleaved and continuous (conventional) disk windings, one single phase transformer,
three different 72 kV capacitor type bushings, and one 72 kV oil type bushing were taken
as test objects and examined through off-line and on-line FRA measurement setups.
8.6.1 Case Study 1
Since the focus of this experiment is on bushing effects, the transformer winding was not
energized. Hence, the transformer bushing tap was directly connected to the FRA
measuring instrument probe without any protection system (see Fig. 8.6). Then, the
conductor of 72 kV capacitor bushing was connected to 66 kV interleaved winding. This
was the same interleaved winding used in previous Chapters as the test object. Figure 8.7
shows a close view of test setup.
Figure 8.6. On-line FRA setup for a transformer interleaved winding.
181
Figure 8.7. Close view of test setup.
8.6.2 Test Procedure
At first, the frequency response of the transformer interleaved winding was recorded
directly through its terminals without bushing (end-to-end measurement). The winding
outer conductors for each disk were accessible for making connection. To have more data
on different conditions of the test object, disk-to-disk short circuits were deliberately
carried out to simulate two different winding defects: one by shorting the outer
conductors of disks 5 and 7 and the other by shorting the outer conductors of disks 5 and
9. The frequency response spectra for normal and faulted cases are shown in Fig. 8.8.
These results are going to be compared to similar measurements through the on-line
setup next.
The on-line FRA test setup was configured according to Fig. 8.6. The frequency response
trace for the winding with connected bushing was recorded. To compare on-line and off-
line FRA spectra, the same disk-to-disk short circuits were made and the corresponding
frequency response spectra were obtained. Measurement results for the on-line setup are
shown in Fig. 8.9.
182
-10
-20
Amplitude [dB]
-30
-40
-50
-60 Reference
Discs 5 and 7 are shortend
Discs 5 and 9 are shortend
-70 2 3 4 5 6
10 10 10 10 10
Frequency [HZ]
Figure 8.8. Frequency response traces of normal and defected winding, off-line setup.
-10
-20
-30
-40
-50
Amplitude [dB]
-60
-70
-80
-90
-100 Discs 5 and 7 are shortend

Reference
-110 Discs 5 and 9 are shortend
2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 8.9. Frequency response traces of normal and defected winding, on-line setup.
8.6.3 Interpretation
RLCM network can be used to model a transformer winding. The FRA test setup in Fig.
4.2(a) is re-drawn in Fig. 8.10(a). The transformer winding in Fig. 4.2(a), has been
replaced by its RLCM network in Fig. 8.10(a) to facilitate physical interpretation.
According to Fig. 8.8, low frequencies are affected mainly by transformer winding self-
183
inductance (Lm). As the frequency increases from 20 Hz to 18 kHz, the inductive reactance
of the transformer winding increases accordingly. Therefore, based on equation (2.7), the
reference frequency response trace will experience a falling trend.
When disks 5 and 7 are short-circuited the self-inductance will change. The short circuit
causes the number of turns in the winding reduced, and consequently the self-inductance
is decreased, see equation (5.8). Therefore, the minimal peak at 18 kHz will be shifted to
higher frequencies according to equation (5.20).
In addition, the minimal peak amplitude takes a lower value due to lower self-inductance
value. A lower value for the self-inductance can be justified using (2.7) and (5.7).
Moving from the first minimal peak towards higher frequencies, the trace starts to
fluctuate. These fluctuations increase when the two disks are short-circuited. In fact, due
to the short circuit, some of turn-to-turn and disk-to-disk capacitances are eliminated and
the series capacitance in transformer winding is reduced significantly. This in turn will
result in winding series capacitance reduction in the interleaved winding as per equation
(3.32), and more oscillations in mid-frequency band, see equation (4.21).
50 ohms
Vin
R-L-C-M
FRA
Winding
Vout
50 ohms
(a)
50 ohms 50 ohms
Cb Cb
Vin
Vin
Bushing Bushing
FRA
FRA
R-L-C-M L
Vout
Vout
Winding Self-inductance
50 ohms 50 ohms
(b) (c)
184
50 ohms 50 ohms
Cb
Vin
Cb
Vin
Bushing Bushing
FRA
FRA
M-Cs-Cg
Mutual-inductance Cg
Vout
Vout
Series and shunt
Shunt capacitances
capacitances
50 ohms
50 ohms
(d) (e)
Figure 8.10. FRA test setup behaviour with and without bushing connection, (a) Common FRA test setup, (b)
FRA test setup with bushing connected, (c) FRA test setup with bushing connected (low-frequency
behaviour) (d) FRA test setup with bushing connected (mid-frequency behaviour), (e) FRA test setup with
bushing connected (high-frequency behaviour).
In the case of on-line FRA spectrum interpretation, the FRA test setup with bushing
connection is shown in Fig. 8.10(b). This setup is similar to what illustrated in Fig. 8.6,
where Cb represents the bushing equivalent capacitance. Figures 8.10(c), 8.10(d) and
8.10(e) show the FRA test setup for low-, mid- and high-frequency regions, respectively.
The winding behaviour in low frequencies is modelled through the self-inductance, in mid
frequencies through the turn-to-turn mutual inductance as well as series and shunt
capacitances and for high frequencies by shunt capacitances. According to the spectra in
Fig. 8.9, in very low frequencies the test setup experiences the bushing capacitance as well
as self-inductance of the transformer winding (see Fig. 8.10(c)). The inductive reactance of
the winding self-inductance is less than the capacitive reactance of the bushing
capacitance. Thus, the overall test setup reactance is quite capacitive. This means that the
measurement setup acts as a high pass filter and cannot convey the low-frequency data of
the input signal. Therefore, the low-frequency information of transformer winding is
replaced by the highly oscillating signals such as noise in the trace.
Moving from very low frequencies (less than 1 kHz) to low-frequencies band makes the
winding inductive reactance to become comparable to the bushing capacitive reactance.
The interaction between them produced a resonant point around 12 kHz. Thus, the
frequency response spectrum reaches a turning point and starts a falling trend.
Afterwards, on-line and off-line spectra are quite similar. However, the signal magnitudes
[dB] are different. The discrepancy in magnitude can be interpreted through Figures
8.10(d) and 8.10(e). Indeed, a capacitive reactance is added to the entire circuit causing
change in FRA magnitude.
185
As for the shorted winding, similar behaviours with different resonant frequencies are
observed. This means that the bushing capacitance shows similar influence for different
circumstances.
8.6.4 Case Study 2
In order to find a solution to reach maximum information in on-line setup, further studies
were conducted through replacing the 72 kV capacitor bushing with other 72 kV capacitor
bushings of different capacitive reactance. This study was again conducted on the previous
test object. Figure 8.11 illustrates the reference frequency response spectrum of the
interleaved winding through the off-line measurement setup as well as frequency
response spectra in on-line setup using different capacitor bushings. In Fig. 8.11, the first
bushing has the minimum capacitance value (lowest trace in Fig. 8.11). The capacitance
value of the second bushing is greater than the first one and much less than the third one.
0
-20
-40
Amplitude [dB]
-60
-80
Reference
-100
Measurement through thrid bushing
Measurement through second bushing
Measurement through first bushing
-120 2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 8.11. Frequency response traces of normal and defected winding, off line and on-line setup.
According to Fig. 8.11, maximum deviation from the reference trace occurs when the
transformer bushing has less capacitance value (i.e. maximum capacitive reactance). The
on-line FRA trace becomes closer to the reference trace when the transformer capacitive
bushing has greater value. In this case, the on-line FRA spectrum is completely compatible
with the reference trace above 8 kHz as the transformer bushing has significant
capacitance value. In addition, transformer bushing with higher capacitance value shows
closer magnitude to the original trace. Undesirable high frequency oscillations in very low
frequency regions are eliminated considerably and more information can be extracted. On-
186
line FRA study on various capacitor type bushings showed that higher bushing capacitance
will have less influence on the on-line FRA trace especially in low frequencies.
8.6.5 Case Study 3
In order to study the oil type (conventional) bushing, the capacitor bushing in Fig. 8.6 was
replaced with a 72 kV oil type bushing. A coupling capacitor sensor was designed and
fabricated to play the role of bushing tap. The coupling capacitor was a thin layer of
aluminium strip, wrapped on the porcelain surface of the oil type bushing and isolated
from the ground. Coupling capacitor is usually employed for partial discharge (PD)
measurement as an output sensor. In this study, it was employed as the input terminal.
Figure 8.12 illustrates the reference and recorded frequency response spectra of the
previous test object through the off-line and on-line setups.
0
1
-10
-20
-30
2
-40
Magnitude [dB]
-50
-60
-70
-80
-90
-100
1 Reference
-110 2 Measurement through one coupling capacitor
2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 8.12. Frequency response traces of interleaved (off-line setup and on-line setup through a coupling
capacitor).
According to Fig. 8.12 (graph 2), very high frequency oscillations are experienced in the
low frequencies. These oscillations are present up to 9 kHz. Moving to higher frequencies,
a resonant point due to interaction between the coupling capacitor and the self-inductance
of the winding can be seen.
The characteristic in the remaining of the trace is quite identical to the reference trace. In
addition, the trace achieved through the on-line setup (coupling capacitor) displays
greater absolute magnitude compared to the reference trace over the entire spectrum.
187
Previous study (case study 2) on various types of capacitor bushings revealed that higher
bushing capacitance will have less effect on the FRA trace in on-line setup. Therefore,
coupling capacitor value should be increased to achieve maximum information and
minimum discrepancy. Hence, a new technique is introduced here to increase the coupling
capacitor value. In this technique, several coupling capacitors are mounted on the
porcelain surface of the oil bushing as illustrated in Fig. 8.13.
Figure 8.13. Paralleled coupling capacitors on conventional bushing (oil type bushing).
The bushing conductor (main rod) is considered as the common plate between all
coupling capacitors. The other sides of the coupling capacitors can be connected together
to provide parallel combination. It is worth noting that between the coupling capacitor
plates and the bushing surface on each plate, there is a rubber insulation to isolate each
capacitor plate from the bushing surface and prevent any kind of short-circuit and
decreasing the creepage distance of the porcelain insulator. This configuration will
increase the total coupling capacitance significantly. The frequency response spectra using
five and nine paralleled coupling capacitors are plotted in Fig. 8.14. For comparison, this
figure also shows the reference trace as well as the recorded spectrum for one coupling
capacitor (tested on the interleaved winding).
According to Fig. 8.14, the frequency response trace through on-line setup was
considerably improved when the coupling capacitor value became significant. Among all
traces, the second one in Fig. 8.14 shows the closest spectrum having minimum
discrepancy to the reference trace. This was achieved when nine coupling capacitors were
employed in parallel. Based on Fig. 8.14, high frequency oscillations in the low frequency
band of FRA trace were limited to lower frequencies when maximum coupling capacitors
were employed. The benefits and drawbacks of this technique will be discussed latter.
188
-10
1
-20
-30 2
-40
3
Magnitude [dB]
-50
4
-60
-70 High frequency oscillations

Start point
-80
-90
-100 1 Reference
4 Measurement through one coupling capacitors
-110 3 Measurement through five paralleled coupling capacitors
2 Measurement through nine paralleled coupling capacitors
2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 8.14. Frequency response traces of interleaved winding (off-line setup and on-line setup through
coupling capacitor).
8.6.6 Case Study 4
In order to study other transformer winding types, a 66 kV continuous disk winding was
put to test. This particular winding has 72 disks and 8 single strand turns per disk. The
frequency response traces of continuous disk winding for off-line and on-line setup
through coupling capacitor are depicted in Fig. 8.15.
0
1 Reference
4 Measurement through one capacitor
-10
3 Measurement through five paralleled capacitors
2 Measurement through nine paralleled capacitors
-20
-30
1
-40
Magnitude [dB]
2
-50
3
-60
4
-70
-80
-90
-100
-110
2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 8.15. Frequency response traces of continuous disk winding (off-line setup and on-line setup through
189
According to Fig. 8.15, the same scenario as in the interleaved winding also occurred for
the continuous disk winding. Oscillation trends in the mid-frequency band of the reference
trace spectrum remained unchanged with the on-line FRA setup. It is again obvious that
the increase of coupling capacitance value brings the on-line FRA results closer to what
was recorded through the off-line setup.
8.6.7 Case Study 5
A single phase 11/0.25 kV, 25 kVA transformer was used as another test object to explore
on-line FRA setup impacts upon the frequency response trace. This transformer has oil
type bushings. Hence, an appropriate external coupling capacitor was constructed and
mounted on the porcelain HV bushing.
At first, the reference frequency response trace of HV side was measured through the off-
line FRA setup. Then, the frequency response through one coupling capacitor was
recorded. After that, paralleled combinations for two and three coupling capacitors were
configured and frequency responses were measured. Figure 8.16 shows measurement
results.
-10 1
-20 2
3
-30 4
-40
Magnitude [dB]
-50
-60
-70
-80
-90
1 Reference
-100 4 Measurement through one coupling capacitor
3 Measurement through two paralleled coupling capacitor
-110 2 Measurement through three paralleled coupling capacitor
2 3 4 5 6
10 10 10 10 10
Frequency [Hz]
Figure 8.16. Frequency response traces of single phase transformer (off-line setup and on-line setup through
According to Fig. 8.16, coupling capacitor increment in on-line FRA test setup appears to
shift the response closer to the reference trace and improve matching. This in turn results
in maximum information achievement.
190
On the other hand, total capacitive reactance decreases when the number of coupling
capacitors increases. Also, it has less influence on frequency response trace oscillations. In
addition, significant capacitive reactance will shift the first minimal peak to higher
frequencies according to equation (5.20).
8.7 Discussion
Off-line FRA provides significant information about transformer condition. On-line FRA
measurement is going to be employed on transformer with the aim to achieve as much
information as possible. This study examines the sensitivity and accuracy of the on-line
measurements when it is conducted through the transformer bushing tap. Since the
bushing imposes a capacitive reactance to the test setup, the trace should be interpreted
by taking the bushing effects into account. Sensitivity and accuracy have been discussed in
the first case study.
On one hand the bushing has crucial effects in very low frequency bands and some part of
the information is missing due to the bushing characteristics. On the other hand,
transformer on-line FRA setup must be designed to avoid power frequency penetration
and superposition effect on the FRA signal in the bushing tap. In fact, a built-in high pass
filter must accompany the paralleled impedance in Fig. 8.3 to avoid power frequency
superposition impacts on the transformer frequency response trace which seems to be
inevitable. In addition, high voltage disturbances in overhead line can be transferred to the
bushing tap and harm the paralleled impedance. Hence, a protection system for this
paralleled impedance needs to be developed to avoid unwanted damage.
According to case studies, on-line FRA measurement through capacitance graded type
bushings will lead to achieving more information as compared to oil type bushings.
Using paralleled coupling capacitors can improve on-line FRA trace and eliminate a part of
undesirable high frequency oscillations. In this circumstance, bushing insulation
coordination will remain as another challenge. In fact, coupling capacitors should be
designed and manufactured based on transformer bushing characteristics to satisfy
insulation coordination. It is recommended to examine the bushing capacitive reactance as
a part of test setup before on-line FRA measurement setup is employed.
It is obvious that the measurement setup for online FRA has more influenced on the very
low frequency part of the FRA trace. However, insulation characteristic changes such as
moisture and temperature do not contribute to this part of spectrum. Therefore, these
changes of transformer could be detectable through online FRA implementation.
191
8.8 Conclusion
In the first part of this Chapter; advanced methods for transformer winding deformation
diagnosis were discussed. The problem in migrating off-line transformer winding
deformation recognition methods to on-line methods was highlighted. In this regard, on-
line frequency response measurement was described and discussed.
In the second part of this Chapter, recommended practical setups for on-line FRA
measurement were elaborated and technical issues associated with the measuring process
were discussed. The possibilities to collect as much information as possible about
transformer condition through the on-line FRA setup were also studied using different
case studies.
Possible solution to get maximum information via on-line setup was introduced and
discussed, and the results through this solution in on-line FRA test setup were examined.
In order to yield traces without interferences over the very low frequency band, it was
recommended to install a built-in high pass filter in the paralleled impedance of the test
setup and determine the cut-off frequency.
In summary, it was discussed that the online FRA measurement is viable solution for on-
line moisture migration recognition from the transformer paper insulation as well as
distinguishing the insulation characteristics impacts through winding deformation on FRA
trace. Further studies are recommended to investigate theoretical and practical challenges
in replacing off-line methods by on-line schemes.
192
Chapter 9.Conclusion and Future Research
Chapter 9
Conclusion and Future Research
9.1 Conclusion
One of the most crucial defects that can put transformers out of service for a long time is
the transformer winding deformation or active part displacement. Over the years, a
number of diagnosis methods have been introduced and employed to detect the mechanical
defects within this valuable asset. Among all, the FRA method is quite capable of providing
worthy information about mechanical integrity as well as electromagnetic behaviour of the
transformer windings. The idea of using FRA as a method for monitoring the condition of
power transformers is not new. It has already been established that FRA spectrum is quite
sensitive to figure out the mechanical defects and the FRA method is widely used
nowadays. However, FRA spectrum interpretation is still under development. In addition,
its sensitivity to the transformer insulation characteristic variations such as the
temperature and humidity is quite significant. This sensitivity can lead to an incorrect
prognosis and concerns on this issue had been raised in a number of Standards.
Since one of the long term goals in transformer condition monitoring is to prevent any
incorrect decisions in transformer diagnosis, the intention of the work presented in this
thesis was to study and highlight the discrepancy between the mechanical defect and the
impacts of insulation characteristic (temperature and moisture) on the FRA spectrum.
Based on this, the preliminary step was to review of the main reasons of winding
deformation and specifically discussed the effects of electro-dynamic forces caused by
short circuit currents. The FRA method and its circuit setup were introduced. It was
demonstrated that FRA is superior to other available methods such as SCI in terms of
winding deformation recognition. Statistical indicators as the available evaluation
methods for frequency response interpretation were also discussed in detail.
Afterwards, using turn-to-turn inductance calculation the self- and mutual-inductances of

the transformer winding were analytically derived and the inductance matrix was
193
achieved. Series and shunt capacitances of transformer winding were then derived
through similar approach. An air-core glassy model transformer was designed and
manufactured and used as a test object. The analytical calculation on winding parameters
was verified using practical measurement on this object. The approach developed here can
potentially help other researchers studying transformer winding through the detailed
model.
Having obtained an air-core winding model through the travelling wave and transmission
line theories, a detailed analysis and interpretation for mid-frequency oscillations in FRA
spectrum was provided. The reason of resonances and anti-resonances in the mid-
frequency band of FRA spectrum was clarified through the mathematical calculations. FRA
mid-frequency oscillations dependency on the inductance as well as series and shunt
capacitances was explored and the simulation was carried out to examine this achievement.
The glassy model transformer was utilized to compare the simulation and measurement
results. This research revealed that a winding with small impulse voltage distribution
coefficient (α) will result in more oscillations in the mid-frequency region of FRA trace while
greater α will give a steady trend in this frequency band. It was also demonstrated that the
influence of transformer metal core would be negligible in terms of FRA mid-frequency
interpretation.
The interpretation of the low-frequency region in FRA trace was also discussed. This study
showed that the first minimal peak in FRA spectrum comes through the transformer middle
limb impact while the second one is influenced by the lateral limb. In order to validate this, one
of the winding frequency response trace of a power transformer was generated through other
deviated traces. Furthermore, according to achievements in this study similar equivalent
magnetic circuits for a given FRA setup will lead to identical FRA spectra. This concern was
studied through practical measurement. This in turn results in a new procedure developed for
transformer core defect recognition. In addition, the impact of the shunt capacitance on the
first anti-resonance in FRA spectrum was studied and discussed through mathematical and
practical approaches. It was discovered that not only the shunt capacitance can shift the
minimal peak position in the low-frequency region; it also has considerable effect on the
frequency response trend in the approach to the first anti-resonance in FRA spectrum.
In terms of winding mechanical defects, the study on axial and radial deformations in
transformer winding was rigorously conducted through the analytical approach. This was
then completed through numerical examples as well as simulations. It was found that the
transformer winding inductance will be changed due to the axial deformation in a winding
194
disk. In addition, radial deformation will change the shunt capacitance of the winding
significantly.
Since the frequency response of transformer winding is completely dependent on its

parameters, the FRA spectrum deviation due to changes in the winding parameters was
also studied. In this case, simulation studies were verified by practical measurements on
different objects. Particular influences of the various parameters (winding inductance,
series and shunt capacitances, resistance, conductance to ground, and turn-to-turn
conductance) on FRA trace were discovered and clarified in this thesis.
The influence of temperature and moisture content on the frequency response spectrum
has recently been a topic of vigorous discussion among researchers. This thesis indicated
that the transformer temperature and moisture variation can lead to FRA spectrum
deviation. In fact, moisture migration from the paper into the oil insulation will cause the
FRA spectrum shifting horizontally (frequency-axis) to lower frequencies, whilst moisture
diffusion from the oil to the paper insulation will shift the FRA spectrum to higher
frequencies. Detailed study on moisture migration for different test objects suggested that
changes of 0.5 % moisture in paper insulation will move the local FRA resonance and anti-
resonance peaks by 0.79 %.
Based on the achievements in this thesis, the FRA method appears to be able to provide
significant information on moisture migration from the paper insulation. It may be utilized
as an effective tool for double–checking the efficiency of transformer dry-out process.
Thus, an FRA measurement could be performed before and after a transformer dry-out
operation to ensure satisfactory moisture diffusion.
Study on the statistical indicators for different temperatures and moisture contents in
transformer winding revealed that they are quite susceptible to these effects and thus can
lead to incorrect prognosis. Indeed, available indicators should be modified or their
existing criteria must be revised to distinguish winding deformation from insulation
characteristic impacts on FRA trace. This was discussed in detail and a method was
recommended for the first time to address this issue.
Apart from the recommended method, a possible solution to distinguish winding

deformation from insulation characteristic effects on FRA spectrum is by implementing
on-line FRA measurement. In reality, insulation deterioration occurs gradually over a long
time scale and so the FRA spectrum is expected to also change gradually. On the other
hand, winding deformation will trigger a sudden shift of the FRA graph. On-line FRA
measurement monitors the winding frequency response continuously, and thus is able to
195
distinguish the gradual change of FRA spectrum due to the insulation characteristic
variation from the fast displacement of FRA trace due to the winding deformation.
Undoubtedly, more work is necessary to establish a robust technique. To open the doors
for future study, the online FRA and its setup was discussed.
9.2 Future Research
Based on the findings in this study, there are a number of subjects that can be targeted for
future research.
9.2.1 FRA Test Setup Development
To date, different test setups have been recommended to measure the frequency response
of the transformer winding. These setups were discussed in detail in this study.
The main goal of each different FRA test setup is to recognize the winding deformation;
however, they are also capable of checking the transformer core integrity. As a result
discovered in this thesis, detailed investigation on core integrity is feasible by adding some
extra wiring into the available FRA setups. This can be achieved through additional extra
short circuit connection on predetermined open terminals of a specific winding. This
work provides the opportunity to localize the fault in the transformer core. Note that the
additional short circuit is not required to be performed on all open winding terminals,
synchronously. It should be implemented step by step for different open winding
terminals as discussed in Chapter 5. Indeed, deliberate short circuit on the terminals of a
winding will impede the flux to flow through that winding. This work helps to direct the
flux flow through a specific core leg and concentrate on the fault recognition in that area.
Swapping deliberate short circuit in the setup from one terminal-pair to the others can
narrow down the fault investigation and lead to fault localization. Hence, rather than using
the available setups for FRA measurement, a study focused on different FRA test setups
would be quite valuable.
9.2.2 Transformer Humidity Recognition Using FRA
The moisture influence on the insulation aging of the transformer is quite obvious for all
utility engineers as well as transformer researchers. Having continuous information about
the humidity variation in the paper insulation is highly desirable. Therefore as discussed
in this study, different methods including off-line and online, direct and indirect
techniques have been employed for moisture recognition in transformer for many years.
These methods are able to provide absolute value of the moisture content in the paper
insulation [Appendix F]. However, the implementation of these methods requires that the
196
moisture equilibrium is achieved between the paper and oil insulations. Otherwise, the
results will not be accurate. On the other hand, whenever the transformer load is changed,
the moisture migrates from one phase (solid/liquid) to the other phase due to the
temperature variations. Thus, new equilibrium should be achieved for moisture
measurement. To reach the new equilibrium, the equilibrium time must be reached.
Transformer equilibrium may take several hours or even much longer. It also depends on
insulations’ aging as well as their contaminations. Hence, real-time measurement of
moisture content in transformer paper insulation is not available at present.
The study in this thesis revealed that FRA signature is quite sensitive in detecting moisture
migration from the paper into the oil insulation and vice versa. Although FRA sensitivity
cannot disclose the absolute value of the moisture in paper insulation, it is able to show
the migration trend. Having obtained the moisture migration trend through FRA and also
the initial condition of FRA signature as well as the initial value of moisture content; it
could be then feasible to recognize the moisture content in paper insulation. Indeed,
perfect implementation of online FRA measurement can change this hypothesis to a real-
time application. Additional research is required to reach real-time humidity recognition
using FRA.
9.2.3 Transformer Dry-out Assessment Using FRA
Transformer dry-out is required after its manufacturing, commissioning, prolonged storage

without nitrogen, maintenance, detection of high moisture content or internal inspection.
As discussed earlier, different methods could be employed for transformer drying. To
evaluate transformer moisture content after the dry-out process, Frequency Dielectric
Spectroscopy (FDS), Karl-Fischer Titration (KFT), paper sample method (Dean-Stark) or
Dielectric Dissipation Factor (DDF) measurement could be implemented. In addition, DDF
measurement over a wide frequency range which is quite similar to FDS has been used to
provide real-time information on the insulation's water content during the actual drying
process. The research in this thesis showed that the FRA spectrum is quite sensitive to
deviation from the reference value due to the moisture variation in paper insulation. This
deviation can be exploited to evaluate the efficacy of the dry-out process. In order to reach
an acceptable level of satisfaction, additional research on this recommendation is
required. Extensive study on different test objects as well as study on insulation
parameters of wet and dry transformers may lead to an industry-accepted method to
assess transformer dry-out process. Different measurement implementations and building
197
a comprehensive database can help significantly and should be considered as the next
move.
9.2.4 Oil and Paper Insulation Aging and Oil Replacement
The influence of temperature and moisture variation on the frequency response spectrum
was studied in this thesis. These two insulation parameters seem to cause the most
impacts on FRA spectrum. However, it would be useful to investigate the influence of other
parameters such as oil and paper insulation aging as well as oil acidity, oil interfacial
tension, oil contamination, oil viscosity, oil breakdown voltage. In addition, sometimes the
oil condition is such that it requires replacing the existing oil with new oil, or maybe a
mineral oil is changed using alternative insulating liquids, these circumstances should be
studied in detail and their impacts on FRA spectrum must be explored. The results of such
studies may then lead to some more considerations in the setup of FRA signature
measurement in future.
9.2.5 On-line Transformer Winding Deformation Recognition
Since smart high voltage monitoring systems are under development now, it is required
that all of off-line diagnosis measurements are able to be upgraded to perform on-line.
Among all, FRA has unique potential to be implemented on-line and provide real-time
information on transformer active parts. Off-line FRA measurement suffers from a number
of drawbacks: costly service outage for testing, de-energising and disconnecting the
transformer from the bus bars, conducting the test with its setup in similar conditions
(temperature, moisture content, cable connection) to the previous FRA signature
measurement. All these limitations motivate researchers to study the implementation of
on-line FRA as the future work.
198
Appendix A. Developed Software to Calculate Statistical Indicators
Appendix A
Developed Software to Calculate

Statistical Indicators
A.1 Introduction
A common method to interpret the frequency responses of transformer windings is using

statistical indices or indicators which have been introduced by various researchers over
the years. These indicators provide information on winding deformation of transformer.
The criteria to explore winding deformation have been determined for some of them,
while others are still under investigation. Among those methods with criteria provided,
some have been implemented in available FRA equipment in industry as a module to
assess FRA spectrum. It means that whenever the FRA test is conducted, the
technician/expert can easily import the reference and measured spectra to the developed
module and compare them together. The output would be then a colour indicating
transformer winding condition. Demonstrating red, yellow and green colours as the
outcome of the implemented method in turn means that the transformer winding is
severely, slightly deformed or may be it has been remained unchanged, respectively. Since
different manufacturers used different methods in their own developed software, this
study has tried to develop a software package to calculate most or perhaps all the
available statistical indicators. This work provides the opportunity to compare and assess
all the criteria together and make more accurate decision on transformer condition. The
idea of this work was introduced and planned by the author for the first time, and then set
as a project for other students to implement the coding. Subsequently, the developed
programs were debugged by the author, and substantial improvement was made in
reaching the current form. Detailed information on the developed software is provided in
this Appendix.
199
A.2 Implemented Statistical Indicators
A.2.1 Correlation Coefficient (CC)
The correlation coefficient method was given by (2.8) in Chapter 2. It is well-known that
this method is not sensitive to changes in the shape of the responses characterized by
variation in the magnitude. In fact, it is more sensitive to generation, elimination, and also
deviation in existing resonant frequencies [20].
Different criteria for CC have been introduced in the literatures, but implemented criteria
for developed software are based on that presented in Chapter 2.
A.2.2 Maximum Absolute Difference (DABS)
The maximum absolute difference was given by (2.9) in Chapter 2. It is well-known that
this method is sensitive to slight differences in amplitude between the FRA responses. A
table provided by Secue et al [20] discusses the characteristic of DABS.
A.2.3 Absolute Sum of Logarithmic Error (ASLE)
The formula for the ASLE is:
i1 20 log
Ns
10 Yi  20 log10 X i
ASLE( X ,Y )  (A.1)
Ns
where Xi and Yi are the ith elements of the fingerprint and measured FRA traces
respectively, and Ns is the number of elements (or samples). “ASLE were proposed by the
authors in order to correct these undesirable characteristics of the CC and SD. ASLE was
presented as the most reliable parameter which was designed to make the fully log-scaled
comparison in the magnitude frequency response; its application considers a previous
process of interpolation proposed by the operator. The normal range of variation for these
parameters has not been set yet” [20].
A.2.4 Min-Max (MM)
The Min-Max method was given by (2.10) in Chapter 2. If the resonance magnitude is not
changed; MM would not be sensitive to deviation of resonant points [20]. In [20], it is also
stated that MM is quite sensitive to the changes of the magnitude variations in FRA
spectrum as a consequence of new resonant frequencies creation or the elimination of
existing resonant frequencies, or even shifting in them.
200
A.2.5 Standard Deviation (SD)
Standard deviation was discussed in Chapter 2. Its equation was provided in (2.11) and
criteria were given in Table 2.2. Standard deviation shows how much variation exists from
the average. A low value for standard deviation shows that the data is very close to the
mean or reference value, whereas high value indicates that the data points are spread out
over a large range.
A.2.6 Spectrum Deviation (σ)
Spectrum deviation was discussed in (2.12). The spectrum deviation was first proposed as
a statistical indicator for comparing FRA traces by Bak-Jensen et al [149], and then became
popular.
A.2.7 Cross Correlation Coefficient (CCF)
The Cross Correlation Coefficient is given by:
i1 X i  X Yi  Y 
Ns
CCF( X ,Y )  (A.2)
i1   i1  
N 2 N 2
Xs
i  X Yi  Y s
where and are the arithmetic average for {Xi}, {Yi} and n=1… Ns. In terms of CCF
interpretation, cross-correlation takes two sets of numbers and looks at how similar they
are. If assuming two series of numbers such as from two FRA spectra are perfectly or
nearly matched, the CCF obtained is close to 1.0. If they have absolutely no correlation, in
other words are completely random, the CCF would be 0. “If assuming both spectra are
related diametrical, they would have a negative CCF. In FRA analysis negative CCF are not
common, but they do occur on occasion. Regardless, negative correlation coefficients are
not considered acceptable when trying to look for deviations between traces” [26]. In [26],
it is highlighted that “inappropriately assigned region boundaries, or single resonance
shifts for example, could cause only a minor change to the CCF and in truth indicate a
substantial problem”.
A.2.8 Relative Factor (R)
This method which is basically coming from Chinese standard [74] is based on calculation
of the covariance for FRA spectra. Firstly, the standard variance for reference and
measured spectra should be calculated [25]:
N s 1 2
 N s 1 
  i 0
1 1
DX   X i  X i  (A.3)
Ns i 0  Ns 
201
N s 1 2
 N s 1 
  i 0
1 1
DY   Yi  Yi  (A.4)
Ns i 0  Ns 
Then:
N s 1 
 
2 2
 N s 1   N s 1 
  i 0  i 0
1 1 1
C XY   X i  X i   Yi  Yi   (A.5)
Ns  N Ns  
i 0   
s

Normalization of covariance factor is given by [25]:
C XY
LRXY  (A.6)
DX DY
Calculation of Relative factor (Rxy) is then given by [25]:
10 , 1  LRXY  1010 

RXY    (A.7)
 log 1  LRXY  , Otherwise
  
Next, the introduced criteria in Table (2.3) should be taken into account to interpret FRA
responses through the Relative factor.
All above mentioned methods have been used to develop a comprehensive software tool
to calculate statistical indicators. This product is believed to have strong potential to be
commercialised.
202
indicators.
Figure A.1 shows a typical screen display of the developed software to calculate statistical
A.3
Developed Software
SFRA
Amplitude-.Piot
Figure A.1. Screenshot of developed software to calculate statistical indicators.
- - Mearurement
- - Reference

~7---~~~~~~~~-~-~~~~~~~-~-~~~~~~--~-~~~~~~--~-~~~~~~--~~-~~~ww
~ ~ ~ ~ ~ ~ ~
frequeocy(Hz)
Ampl1tude·Piot
,-~·
l~: ::::::
~ ~
•: : : : •:::: : ~ ·:= : ~'~1 ~
------,- Statistgl Oata

~
frequency(Hz)
~
~~ ~
~
1
II "'"""'""' I TheniJITWrol frequency ~rm( 1 ,2,3 or • I Fui FrequencyRan90 cc CCF DABS

""
.., SD SI3IIA
PUASE SET THE CRITEFM.:
Good
The IIUI!t:let ol aqilude.pl\ase or npedance ~rm( 1 ,2,3 or 4)
2007 • 1 98902t..OOS 093112 0949387 l.38tS4 t 2068 t 29574 10 503003 0132619
(...... )
I Meuw~Ne I The IMHTtJer of frequencycollrm(1,2,3or4) Use-rSelf-defint

~~
War;r~al
ThtiiUll'tlef'ofall1)flude,pl'laseor~nce~rm(t ,2,3or •)
SIJbFrequencyRan90 cc CCF DABS
""
, SD .. SI3IIA
- w)
..........
(«d)
~
Grapha- - - - - ;
~
Frequency Uni(MZJ 2007
PEASE 00 NOT EXCEED FRfOUEHCY UlrT

1989018.917
Q ~
GJEJBG
WHEN APPlYHG SUB FREQUENCY RANGE
l.i'tearxctlt
118BG B
203
Appendix B. Tables and Formulas for Inductance Calculation
Appendix B
Tables and Formulas for Inductance

Calculation
B.1 Introduction
In order to calculate the winding inductance in Chapter 3 and winding inductance

variation in Chapter 6, this Appendix provides the related formulas as well as tables. It
should be emphasized that pre-calculated tables here have been taken from ‎[85], and
summarized for convenience. In addition, the analytical approach conducted in the thesis
on self- and mutual-inductance calculations under buckling is given in this Appendix.
These calculations are given for the first time, and nonlinear integrals obtained from the
calculations are solved using the MATLAB software.
B.2 Tables and Formulas
Table B.1. Values of KN for single layer coil ‎[85].
h/2R K h/2R K h/2R K h/2R K

0 0 0.10 0.203324 0.20 0.319825 0.30 0.405269
0.01 0.034960 0.11 0.217044 0.21 0.329479 0.31 0.412650
0.02 0.061098 0.12 0.230200 0.22 0.338852 0.32 0.419856
0.03 0.083907 0.13 0.242842 0.23 0.347960 0.33 0.426890
0.04 0.104562 0.14 0.255011 0.24 0.356816 0.34 0.433762
0.05 0.123615 0.15 0.266744 0.25 0.365432 0.35 0.440474
0.06 0.141395 0.16 0.278070 0.26 0.373818 0.36 0.447036
0.07 0.158119 0.17 0.289019 0.27 0.381986 0.37 0.453450
0.08 0.173942 0.18 0.299614 0.28 0.389944 0.38 0.459724
0.09 0.188980 0.19 0.309876 0.29 0.397703 0.39 0.465860
0.10 0.203324 0.20 0.319825 0.30 0.405269 0.40 0.471865
Pm ( )  (2m 1)(2m  3)
m! (B.1)
 m
.   m( m  1)  m 2  m( m 1)(m  2)(m  3)  m4  ... 

 2m(2 m  1) 2.4(2m 1)(2m  3) 
204
m Pm ( )  mPm1( )
Pm ( )  (B.2)
 2 1
Table B.2. Values of R0 for inclined circles ‎[85].
μ β=0 β =0.1 β =0.2 β =0 β =0.1 β =0.2

0 0.4099 0.4741 0.6078 0.5300 0.5660 0.6550
0.1 0.4114 0.4763 0.6095 0.5330 0.5680 0.6540
0.2 0.4161 0.4807 0.6150 0.5360 0.5730 0.6580
0.3 0.4229 0.4890 0.6246 0.5440 0.5780 0.6670
0.4 0.4311 0.5012 0.6389 0.5560 0.5920 0.6800
α=0.9 α=0.8
0.5 0.4472 0.5185 0.6593 0.5730 0.6110 0.7000
0.6 0.4673 0.5431 0.6886 0.5980 0.6330 0.7290
0.7 0.4969 0.5794 0.7308 0.6270 0.6680 0.7650
0.8 0.5433 0.6383 0.7951 0.6800 0.7190 0.8180
0.9 0.6278 0.7313 0.9064 0.7640 0.8070 0.9050
1.0 1 1 1 1 1 1
μ β=0 β =0.1 β =0.2 β =0 β =0.1 β =0.2

0 0.6276 0.6498 0.7104 0.7149 0.7306 0.7746
0.1 0.6291 0.6513 0.7121 0.7164 0.7321 0.7761
0.2 0.6337 0.6562 0.7172 0.7209 0.7366 0.7806
0.3 0.6420 0.6645 0.7259 0.7287 0.7444 0.7883
0.4 0.6542 0.6769 0.7387 0.7403 0.7560 0.7996
α=0.7 α=0.6
0.5 0.6714 0.6943 0.7566 0.7563 0.7718 0.8150
0.6 0.6950 0.7182 0.7807 0.7779 0.7932 0.8354
0.7 0.7279 0.7512 0.8135 0.8071 0.8218 0.8621
0.8 0.7753 0.7984 0.8585 0.8472 0.8608 0.8972
0.9 0.8503 0.8710 0.9217 0.9055 0.9160 0.9429
1.0 1 1 1 1 1 1
Table B.3. Values of F for parallel circles, rd/2R=Λ, ‎[85].
cosθ'' Λ=1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
0 -4.053 -1.953 -0.467 0.525 1.145 1.485 1.622 1.621 1.531 1.382 1
0.1 -1.508 -1.023 -0.168 0.550 1.055 1.357 1.495 1.513 1.448 1.327 1
0.2 -0.724 -0.387 0.117 0.585 1.010 1.264 1.392 1.421 1.376 1.278 1
0.3 -0.237 0.013 0.348 0.696 0.989 1.195 1.308 1.341 1.310 1.233 1
0.4 0.101 0.291 0.5246 0.766 0.983 1.144 1.239 1.271 1.252 1.191 1
0.5 0.351 0.493 0.658 0.829 0.984 1.105 1.181 1.211 1.199 1.153 1
0.6 0.544 0.647 0.761 0.878 0.987 1.075 1.132 1.158 1.151 1.115 1
0.7 0.695 0.766 0.842 0.920 0.991 1.050 1.091 1.111 1.108 1.085 1
0.8 0.818 0.861 0.907 0.952 0.995 1.031 1.056 1.069 1.069 1.055 1
0.9 0.917 0.937 0.958 0.979 0.998 1.014 1.026 1.032 1.033 1.026 1
1.0 1 1 1 1 1 1 1 1 1 1 1
205
B.3 Self and Mutual Inductances under Buckling
B.3.1 Biot-Savart Law and Inductance Calculation
One of the most fundamental laws in computation of magnetic field is the Biot-Savart law
which is given by:
 0 J  aR
4 v Rd2
B dv (B.3)
where J is the current density vector [A/m2], v is the volume containing current, and Rd is
the proposed distance. On the other hand Jds = I; therefore, Jdv = Jdsdl = Idl. Hence,
according to (B.3) dB is given by ‎[78], ‎[104]:
0 I
dB  dl  aR (B.4)
4 Rd2
According to Fig. B.1 and also the magnetic field generated through I, the Biot-Savart law
can be explained as:
0 Idl
dB  sin  an , (B.5)
4 Rd2
where, an is the unit vector perpendicular to the plane containing dl and ar, dl an = dl × ar.
dB Rd
aR
θʹ I
dl
Figure B.1. Biot-Savart law.
Based on Gauss’s law for a circular filament having radius Ra and carrying the current I,
the flux density, ϕ, is given by:
Ra 2
   B.ds  (B.6)
s
  B rd dr
0 0
z
where B = Baz and can be obtained through (B.5), and rdϕʹ= dl. Thus for a circular filament
having radius Ra equation (B.5) can be obtained as:
206
0 I (B.7)
dBz  Ra d  sin  
4 Rd2
where, Rd is the distance between the differential current segment and the point of
proposed magnetic field and obtained as (B.8) for a specific point r away from the centre.
θʹ would then be the angle between the differential current vector and the vector directed
from it to the point as shown in Fig. B.2.
Rd2  Ra2  r 2  2Ra r cos   (B.8)
90
120 60
I
θʹ
150 30
Ra αʹ
180
ϕʹ Rd 0
r
B
210 330
240 300
270
Figure B.2. Magnetic flux determination for circular filament, ‎[104] taken and modified.
Hence, equation (B.7) is modified as:
0 IRa
dBz  d  sin   (B.9)
 
2
4 R  R  r  2 Ra r cos  
2
d
2
a
2
θʹ = π/2 + αʹ; thus, sinθʹ = sin (π/2 + αʹ) = cos αʹ. Using the law of cosines ‎[104] r2=Ra2+Rd2-
2RaRd cosαʹ, and therefore ‎[104]:
Ra2  Rd2  r 2 Ra  r cos   (B.10)

sin    cos    
2 Ra Rd Rd
Based on this, Bz(r) is given by:
2
0 IRa Ra  r cos  
Bz (r ) 
4  d  (B.11)
R 
3
0
2
a  r  2 Ra r cos  
2 2
207
In ‎[104], it is stated that the second method for Bz determination is using the vector
magnetic potential as given by:
(rA  ) (B.12)
Bz   A   .
rr
Therefore using (B.7) and (B.12), equation (B.6) for a filament having radius Ra is given by:
a   Bz ds   A dl. (B.13)
sa ca
Apart from the last approach in Chapter 3 for the mutual-inductance calculation through
Grover’s formula, the mutual-inductance between two circular filaments having radii Ra
and Rb are also obtained as:
b (B.14)
M ab  ,
Ia
where ϕb is the induced magnetizing flux on the second loop due to the current initiated by
the first loop. Hence, Mab is given by (see Fig. B.3):
M ab  A
cb
ab dlb
0 1 RR 2 2 cos a

4  dlb .dla  0 a b
4   db da (B.15)
R 
1
ca cb
Rab 0 0 2
 R  d  2 Ra Rb cos a
2 2 2
a b
 
1
[ Note that: Rab  Ra2  Rb2  d 2  2 Ra Rb cos a 2
]
It is again obvious that the mutual-inductance between two circular filaments is only a
function of their shapes as well as orientations. This was stated in ‎[78] and ‎[104].
Rb
x
ϑb
sb
y dlb
d Rab
I
Ra
X
ϑa
sa
Y dla
Figure B.3. Concentric circular filaments.
208
Using Maxwell’s advices in ‎[78], reference ‎[104] is discussed on the solution of (B.15). In
fact, changing the variable ϑa to 2θʺ, cos ϑa = cos2θʺ=2cos2θʺ-1, and dϑa = 2dθʺ will simplify
the equation. Thus, Mab will be calculated as ‎[104]:
0 Ra Rb  2 cos 2a
M ab   da
 R  R   d 
1
2 0
 4 Ra Rb cos a
2 2 2 2
a b
 R R 
1
2  k cos 2a 4 Ra Rb (B.16)
 0 a b  da , k 2 
 Ra  Rb 
1  k 
1 2
2 0 2
cos 2 a 2  d2
2  2

[ Note that: k cos 2a  2 cos 2 a  1    k   1  k 2 cos 2 a ]
k  k
  
Having information about complete elliptic integrals K(k) and E(k), equation (B.15) is
given by ‎[78] and ‎[104]:
 R R 
1
2  k cos 2a
M ab  0 a b  da
1  k 
1
2 0 2
cos a 2 2
 
2  
 0  Ra Rb     k 
1 2
 d

1 1
2 2
 1  k 2 cos 2 a 2
0  k  a
  1  k 2 cos 2   
1
2k
 a 
 

2  (B.17)
 0  Ra Rb  2    k 
1 2
 2

1 1
2
 1  k sin a
2 2
d
0  k  a
  1  k 2 sin 2   
1
2 k
 a 
1  2  2 
 0  Ra Rb  2   k  K (k )  E (k ) 
 k  k 
where:
 
1
1  k 
1
K (k )   2
da , E (k )   2 2
sin 2 a 2
da
1  k 
0 1 0
2
sin a 2 2
Therefore, (B.17) shows the analytical approach to determine the mutual inductance for
concentric circular filaments.
B.3.2 Mutual-Inductance under Buckling
B.3.2.1 Buckling of a Single Filament
Equation (B.15) is utilized to calculate the impact of buckling on the mutual-inductance of

two concentric circular filaments as shown in Fig. (B.4).
209
z
90
120 60
30
150
Rb 0
x
ϑb
180
210
sb 330
240 300
y 270
dlb
d Rab
I
Ra
X
ϑa
sa
Y dla
Figure B.4. Concentric circular filaments, inward buckling demonstration for the second loop.
It should be noted that in the case of inward buckling the winding radius for the span
faced buckling is defined as Rbd = Rb + 0.5rʹ(cos ηθ-1), and outward buckling is
mathematically expressed as Rbd = Rb - 0.5rʹ(cos ηθ-1). η is the ratio of entire trigonometric
circular span (2π) over the deformation span (rad) as illustrated in Fig. 6.4, and r'
represents the deformation radius. Based on this, the mutual-inductance is given by:
0 Ra 2 2 
4 0 0
 
M ab
cos1  Rb  0.5r (cos2  1) 
d2 d1
R 
1
  Rb  0.5r (cos2  1)   d  2 Ra  Rb  0.5r (cos2  1)  cos1 (B.18)
2 2 2 2
a
0 Ra Rb 2 2 cos1

4   d2 d1
 Ra  Rb  d  2Ra Rb cos1 
0 2  1
2 2 2 2
where, the first part of (B.18) comes through the span facing deformation and second part
represents the circular part in the second filament. This integral is quite non-linear and
complex to solve analytically. Therefore it should be addressed numerically using MATLAB
software. However to continue the equation analytically, we can assume that the influence
of Rb on the total value of Rab as the denominator of the first integral is negligible as
compared to Rb influence as the numerator. This assumption is reasonable for the
filaments which are quite far away, but maybe it is not accurate for close loops. All in all,
having this assumption integral in (B.18), Mʹab is obtained as:
210
0 Ra 2 2  ( Rb  0.5r ) cos 1
 
M ab
4   d2 d1
 
0 0 1
Ra2  Rb2  d 2  2 Ra Rb cos 1 2
0 Ra 2 2  0.5r  cos 1 cos2


4   d2 d1 (B.19)
 
0 0 1
Ra2  Rb2  d 2  2 Ra Rb cos 1 2
0 Ra Rb 2 (1   1 ) 2 cos 1

4  d1
 a b 1
0 1
R 2
 R 2
 d 2
 2 R R
a b cos  2
where the first integral in (B.18) has split in (B.19). One step forward, the calculation of
the second filament influence on the Mʹab is given by:
0 Ra ( Rb  0.5r ) 2 cos 1
 
M ab
2  d1
 Ra  Rb  d  2Ra Rb cos1 
0 1
2 2 2 2
0 Ra r  2 cos 1
 
8 0 R 2  R 2  d 2  2 R R cos  12
d1 (B.20)
 a b a b 1
0 Ra Rb (1   1 ) 2 cos 1
  d1
 a b 1
1
2 0
R 2
 R 2
 d 2
 2 R R cos  2
a b
After simplification of (B.20), Mʹab is then given by:
  R r (1  2 ) 0 Ra Rb  2 cos 1
  0 a
M ab   0 d1 (B.21)
8
 
1
 2  Ra2  Rb2  d 2  2 Ra Rb cos 1 2
Comparing (B.21) with (B.16), it is obvious that the term having (1-2π) subtly reduces the
mutual inductance within inward buckling as it takes a negative value. Accordingly, this
coefficient will change to a positive value (1+2π) when the outward buckling occurred for
the filament (see Fig. 6.4).
B.3.2.2 Buckling of both concentric circular filaments
Compared to the previous subsection, analytical calculation of the mutual-inductance once

both filaments are facing deformation is even more challenging. In this case, having
asymmetrical magnetic flux density generated by the first filament on the second one is
considered as the first challenge. Indeed the filament circumference is not circular; thus,
considering two pairs of opposite sides, the horizontal contribution of the magnetic flux
density cannot be completely neutralized. Therefore, B is not fully aligned with az.
The second challenge would be again the calculation of non-linear integrals, analytically. In
fact in spite of analytical approach, numerical calculation is required on the final integrals
211
to reach the exact value of mutual-inductance. All in all, attempts are made in this
appendix to analytically derive near final integrals. This helps to prepare for numerical
calculations, even though such calculations show subtle changes in mutual-inductance for
usual buckling in transformer windings.
Using the same assumption on Rb in the last subsections, the mutual-inductance of the two
concentric circular filaments is modelled as Fig. B.5 and given as (B.22) using (B.15):
z
90
120 60
30
150
Rb 0
x
ϑb
180
210
sb 330
240 300
y 270
dlb
d Rab
120
I 90
60
30
150
Ra
180
0
X
ϑa
210
sa 330
240 300
Y
270
dla
Figure B.5. Concentric circular filaments, demonstration of inward buckling for both loops.
 cos 1  Ra  0.5r (cos(1 )  1)  


 2  2  
 
1
  0    Ra2  Rb2  d 2  2 Ra Rb cos 1 2 d2 d1
M ab
4 0 0
 
  R  0.5r (cos( )  1)  
 b 2 
 2  2 cos 1  Ra  0.5r (cos(1 )  1) 
 0  Rb d2 d1
4 0
 
2  1
Ra  Rb  d  2 Ra Rb cos 1
2 2 2 2
(B.22)
0 2 2  Ra cos 1  Rb  0.5r (cos(2 )  1) 


4  d2 d1
R 
2 0 1
2
a  R  d  2 Ra Rb cos 1
2
b
2 2
0 Ra Rb 2 2 cos 1

4  d2 d1
 
2 2 1
Ra2  Rb2  d 2  2 Ra Rb cos 1 2
Equation (B.22) contains four different integrals as given by (B.23):
  Int 1  Int 2  Int 3  Int 4

M ab (B.23)
The first integral is obtained as:
212
 cos 1  Ra  0.5r (cos(1 )  1)  

 0 2   2   2 .
 
1
4 0 0  a
Int1   R  Rb  d  2 Ra Rb cos 1
2 2 2
d2 d1

  R  0.5r (cos( )  1)  
 b 2 
 0 2   2 
A  Rb  0.5r (cos2  1) d2 d1
4 0 0

 2  2 
 0 
4 0 0
( A( Rb  0.5r )  A0.5r  cos 2 )d2 d1
(B.24)
2 
0 2  0.5r  

4
0 
A ( Rb  0.5r )2 
 
sin 2 
0
d1
(4 Rb  2 r   r ) 0 2 
4 0
 Ad1
2
cos 1  Ra  0.5r (cos(1 )  1) 
where: A 
 
1
Ra2  Rb2  d 2  2 Ra Rb cos 1 2
From calculation of different spans on the first and second filaments, Mʺab is obtained as:
(4 Rb  2 r   r ) 0 2  cos 1  Ra  0.5r (cos(1 )  1) 

Int1 
2 4  d1
R 
0 1
2
a  R  d  2 Ra Rb cos 1
2
b
2 2
 (2 Ra  r )cos1 
 2 
 
1
2   Ra  Rb  d  2 Ra Rb cos 1 
2 2 2
(4 Rb  2 r   r ) 0

2 8 0  r  cos 1 (cos(1 )
d1

 2 
 
1
 Ra  R 2
b  d 2
 2 Ra Rb cos 1
2

  (B.25)
 (2 Ra  r ) cos 1
 
 2 
 
1
 Ra  Rb  d  2 Ra Rb cos 1
2 2 2

 
(4 Rb  2 r   r ) 0 2 
 r  cos(  1)1
 0  2 2 2  d1
2 8
 

1
2 Ra  Rb  d  2 Ra Rb cos 1 2
 
 
r cos(  1)1 
 
 
1
 2 Ra2  Rb2  d 2  2 Ra Rb cos 1 2 
 
Simplification of (B.25) leads to:
213
 
(4 Rb  2 r   r ) 0 (2 Ra  r ) 2   cos 1 d
Int1 
2 8   1 
 
1
 Ra2  Rb2  d 2  2 Ra Rb cos 1 2 
0
 
 
(4 Rb  2 r   r ) 0 (2 Ra  r ) 2   cos(  1)1 d

4 8   1  1 (B.26)

 Ra2  Rb2  d 2  2 Ra Rb cos 1 2  
0
 
 
(4 Rb  2 r   r ) 0 (2 Ra  r ) 2   cos(  1)1 d

4 8   2  1
 
1
 Ra  Rb2  d 2  2 Ra Rb cos 1 
0 2
 
The second integral which concentrates on the interaction of deformation in the first
filament and the circular part in the second filament is given by:
 
 2 
 cos 1  Ra  0.5r (cos(1 )  1)  R d d
2
Int 2  0
4 0 2   2 2 2 b
 
1 2 1
R  R  d  2 R R cos  2

 a b a b 1 
  (B.27)
0 (1   1 ) 2   cos 1  Ra  0.5r (cos(1 )  1)  
 0  2 2 2 Rb d1

 
1
2 
 Ra  Rb  d  2 Ra Rb cos 1
2

0 (1   1 ) 2 

2 
0
Ad1
where, Aʹ has been defined in (B.24) and its solution is similar to (B.26).
The third integral is obtained as:
 
R 2
 cos 1  Rb  0.5r (cos(2 )  1)  d d
2 
Int 3  0 a
4 2  0  2 2 2 1 
 a 1
2 1
R  R  d  2 R R cos  2
 (B.28)
b a b 
(4 Rb  2 r   r ) 0 Ra 2  cos 1

2 4  d1
 
2   1
R 2  R 2  d 2  2 R R cos 
a b a b 1
2
Using Maxwell’s approach, the solution of (B.28) is similar to that calculated in (B.17).
However, the asymmetrical boundary conditions are considered an issue in this case. The
fourth integral is given by:
0 Ra Rb 2  2 cos 1
Int 4 
4   d2 d1 (B.29)
 
2 2 1
Ra2  Rb2  d 2  2 Ra Rb cos 1 2
214
As discussed earlier, the numerical solution of the above mentioned integrals will lead to
the mutual-inductance of the concentric circular filaments with inward buckling in both
filaments. In the case of outward buckling, the positive sign before rʹ should be replaced by
the negative sign.
B.3.3 Self-Inductance under Buckling
The self-inductance of the non-circular filament as shown in Fig. B.6 can be achieved
through (B.6) and (B.7). The self-inductance is given by:
a (B.30)
L
Ia
where, Ia is the current carried by filament, and ϕa is the magnetizing flux generated by Ia.
Based on this, the magnetizing flux density generated through buckling area is then
defined as:
0 I
dBz 
4 Rd2
 Ra  0.5r  cos  1 d  sin   (B.31)
and Rd is obtained as:
Rd2   Ra  0.5r   cos   1   r 2  2  Ra  0.5r   cos   1  r cos   (B.32)

2
90
120 60
150 30
θʹ I
1)
ϑ-
osη
0.5
r ʹ(C αʹ
Ra R a+
Rd
180
ϑ 0
r B
210 330
240 300
270
Figure B.6. Magnetic flux determination for non-circular filament.
Since θ = π/2 + αʹ, therefore, sin θ = sin (π/2 + αʹ) = cos αʹ. Using the cosines law, r is
obtained as:
215
r 2   Ra  0.5r   cos   1   Rd2  2  Ra  0.5r   cos   1  Rd cos  

2
(B.33)
Thus:
R  0.5r   cos   1   Rd2  r 2

2
sin    cos   
a
2  Ra  0.5r   cos   1  Rd
(B.34)
Ra  0.5r   cos   1  r cos  

Rd
Substitution of (B.34) into (B.31), dBz is obtained as:
0 I  Ra  0.5r   cos   1   Ra  0.5r   cos   1  r cos   

dBz  d 
 R  0.5r cos  1  r  2  R  0.5r cos  1 r cos  
3
2
4 2 2
a a
 
 Ra  0.5r   cos   1    Ra  0.5r   cos   1  r cos  
2
0 I   (B.35)
  
4 
 
3
  Ra  0.5r   cos   1   r  2  Ra  0.5r   cos   1  r cos   

2 2 2
 
Therefore, Bz(r) is obtained as (B.36) including the magnetic flux density generated by the
buckling area as well as the circular part:
0 I
Bz (r ) 
4
2  
 Ra  0.5r   cos   1    Ra  0.5r   cos   1  r cos   d 
2
 
  3 
0


 a   a 
 R  0.5r   cos   1 2  r 2  2 R  0.5r   cos   1 r cos   2 
  (B.36)
2
 IR Ra  r cos  
 0 a  d 
4 2 R 2  r 2  2 R r cos   3 2
  a a 
The self-inductance through the non-circular filament as illustrated in Fig. B.6 is then
given by:
216
0 2 Ra  0.5 r  cos  1

 

L 
2 0 0
 

  Ra  0.5r   cos   1  cos    Ra  0.5r   cos   1   r cos   
2

  drd 
  (B.37)
3
   a 
 R  0.5r   cos   1 2  r 2  2 R  0.5r   cos   1 r cos   2
 a


0 2 Ra Ra cos    Ra  r cos   
2
 2  drd .

2  0 R 2  r 2  2 R r cos   3 2
a a 
The solution of (B.37) can be achieved through numerical calculation using MATLAB
software. However, more simplification for such integrals can be evaluated analytically
using the following integrals ‎[104], ‎[150]:
dx 4ax  2b
  ,
 4ac  b  ax
3 1
 ax  bx  c 
2 2 2 2
 bx  c  2
xdx 2bx  4c (B.38)

 
 4ac  b  ax
3 1
 ax 2  bx  c  2 2 2
 bx  c  2
Self- and mutual-inductance variations under buckling phenomenon are quite dependent
on the winding geometry. In fact, the most important parameters to influence the original
value of the inductance are the ratio η of deformation span (rad), and the deformation
radius r'. Nevertheless, according to literatures ‎[102] and also information achieved in this
study, these variations are negligible for usual buckling in routine transformer windings.
B.3.4 Numerical Example
The self- and mutual-inductance variation caused by winding radial deformation was
discussed in the last subsection. It was also stated that the final analytical formulas
developed for self- and mutual-inductance calculation should be solved and addressed
through the numerical analysis. This is done in this subsection on the example provided in
Chapter 6 (see Section 6.4 and Sub-section 6.4.2) using MATLAB software.
Based on Maxwell equations, the self-inductance of a circular turn having 280 mm

outermost radius and 7 mm width is calculated as:
LMaxwell  1.5338  H (B.39)
217
while this value for the same turn using Grover equations (see L33 in (6.12) with the
central turn of the disk having outermost radius 280 mm) is calculated as:
LGrover  1.6507  H (B.40)
If assuming the circular turn has suffered inward buckling such as depicted in Fig. 6.8 (left
side) and Fig. B.6; then, based on the calculation provided in this thesis the self-inductance
for non-circular loop is obtained as (rʹ=50 mm):
L  1.1954  H (B.41)
In the case of mutual-inductance, for a concentric circular turn 17 mm away from the main
loop, MMaxwell is obtained as:
M Maxwell  0.9977  H (B.42)
This value was calculated as (B.43) through Grover’s approach in Chapter 6.
M Grover  0.9969  H (B.43)
Under the discussed buckling, the mutual inductance is obtained as (B.44) when one of the
turns has suffered buckling (rʹ=50 mm):
  0.8740  H
M ab (B.44)
To complete the example, it is supposed that all conductors in the second disk of the model
winding depicted in Fig. (3.1) have suffered inward buckling (rʹ=50 mm); hence the
inductance matrix for the normal and defected winding is obtained as (B.45) and (B.46),
respectively:
Leq(Maxwell ) 
1.6300 1.3680 1.1003 0.9400 0.8523 0.9399 1.0208 1.0662 0.8157 0.7942 0.7598 0.7171 0.6075 0.6341 0.6558 0.6714 
 
1.3680 1.5818 1.3260 1.0644 0.9080 0.9873 1.0319 1.0208 0.7942 0.7875 0.7663 0.7326 0.6101 0.6314 0.6468 0.6558 
1.1003 1.3260 1.5338 1.2841 0.9540 0.9977 0.9873 0.9399 0.7598 0.7663 0.7596 0.7386 0.6073 0.6225 0.6314 0.6341 
 
0.9400 1.0644 1.2841 1.4861 0.9638 0.9540 0.9080 0.8523 0.7171 0.7326 0.7386 0.7318 0.5984 0.6073 0.6101 0.6075 
0.8523 0.9080 0.9540 0.9638 1.4861 1.2841 1.0664 0.9400 0.8523 0.9080 0.9540 0.9638 0.7318 0.7386 0.7326 0.7171 
 
0.9399 0.9873 0.9977 0.9540 1.2840 1.5338 1.3260 1.1003 0.9399 0.9873 0.9977 0.9540 0.7386 0.7596 0.7663 0.7598 
1.0208 1.0319 0.9873 0.9080 1.0664 1.3260 1.5818 1.3680 1.0208 1.0319 0.9873 0.9080 0.7326 0.7663 0.7875 0.7942 

1.0662

1.0208 0.9399 0.8523 0.9400 1.1003 1.3680 1.6300 1.0662 1.0208 0.9399 0.8523 0.7171 0.7598 0.7942 0.8157 

(B.45)
0.8157 0.7942 0.7598 0.7171 0.8523 0.9399 1.0208 1.0662 1.6300 1.3680 1.1003 0.9400 0.8523 0.9399 1.0208 1.0662 
0.7942 0.7875 0.7663 0.7326 0.9080 0.9873 1.0319 1.0208 1.3680 1.5818 1.3260 1.0644 0.9080 0.9873 1.0315 1.0208 
 
0.7898 0.7663 0.7596 0.7386 0.9540 0.9977 0.9873 0.9399 1.1003 1.3260 1.5338 1.2841 0.9540 0.9977 0.9873 0.9399 
0.7171 0.7326 0.7386 0.7318 0.9638 0.9540 0.9080 0.8523 0.9400 1.0644 1.2841 1.4861 0.9638 0.9540 0.9080 0.8523 
 
0.6075 0.6101 0.6073 0.5984 0.7318 0.7386 0.7326 0.7171 0.8523 0.9080 0.9540 0.9638 1.4861 1.2841 1.0644 0.9400 
 
0.6341 0.6314 0.6225 0.6073 0.7386 0.7596 0.7663 0.7598 0.9399 0.9873 0.9977 0.9540 1.2841 1.5338 1.3260 1.1003 
0.6558 0.6468 0.6314 0.6101 0.7326 0.7663 0.7875 0.7942 1.0208 1.0315 0.9873 0.9080 1.0644 1.3260 1.5818 1.3680 
 
0.6714 0.6558 0.6341 0.6075 0.7171 0.7598 0.7942 0.8157 1.0662 1.0208 0.9399 0.8523 0.9400 1.1003 1.3680 1.6300 
218
Leq (Maxwell ) 
1.6300 1.3680 1.1003 0.9400 0.7322 0.8074 0.8816 0.9371 0.8157 0.7942 0.7598 0.7171 0.6075 0.6341 0.6558 0.6714 
 
1.3680 1.5818 1.3260 1.0644 0.7785 0.8510 0.9054 0.9280 0.7942 0.7875 0.7663 0.7326 0.6101 0.6314 0.6468 0.6558 
1.1003 1.3260 1.5338 1.2841 0.8206 0.8740 0.8950 0.8904 0.7598 0.7663 0.7596 0.7386 0.6073 0.6225 0.6314 0.6341 
 
0.9400 1.0644 1.2841 1.4861 0.8427 0.8635 0.8593 0.8939 0.7171 0.7326 0.7386 0.7318 0.5984 0.6073 0.6101 0.6075 
0.7322 0.7785 0.8206 0.8427 1.2364 1.1954 1.0299 0.9371 0.7322 0.7785 0.8206 0.8427 0.6643 0.6613 0.6499 0.6328 
 
0.8074 0.8510 0.8740 0.8635 1.1954 1.2781 1.2361 1.0661 0.8074 0.8510 0.8740 0.8635 0.6841 0.6907 0.6874 0.6755 
0.8816 0.9054 0.8950 0.8593 1.0299 1.2361 1.3200 1.2283 0.8816 0.9054 0.8950 0.8593 0.6947 0.7109 0.7174 0.7137 

0.9371

0.9280 0.8904 0.8939 0.9371 1.0661 1.2283 1.3622 0.9371 0.9280 0.8904 0.8439 0.6971 0.7215 0.7378 0.7443 

(B.46)
0.8157 0.7942 0.7598 0.7171 0.7322 0.8074 0.8816 0.9371 1.6300 1.3680 1.1003 0.9400 0.8523 0.9399 1.0208 1.0662 
0.7942 0.7875 0.7663 0.7326 0.7785 0.8510 0.9054 0.9280 1.3680 1.5818 1.3260 1.0644 0.9080 0.9873 1.0315 1.0208 
 
0.7898 0.7663 0.7596 0.7386 0.8206 0.8740 0.8950 0.8904 1.1003 1.3260 1.5338 1.2841 0.9540 0.9977 0.9873 0.9399 
0.7171 0.7326 0.7386 0.7318 0.8427 0.8635 0.8593 0.8939 0.9400 1.0644 1.2841 1.4861 0.9638 0.9540 0.9080 0.8523 
 
0.6075 0.6101 0.6073 0.5984 0.6633 0.6841 0.6947 0.6971 0.8523 0.9080 0.9540 0.9638 1.4861 1.2841 1.0644 0.9400 
 
0.6341 0.6314 0.6225 0.6073 0.6613 0.6907 0.7109 0.7215 0.9399 0.9873 0.9977 0.9540 1.2841 1.5338 1.3260 1.1003 
0.6558 0.6468 0.6314 0.6101 0.6499 0.6874 0.7174 0.7378 1.0208 1.0315 0.9873 0.9080 1.0644 1.3260 1.5818 1.3680 
 
0.6714 0.6558 0.6341 0.6075 0.6328 0.6755 0.7137 0.7443 1.0662 1.0208 0.9399 0.8523 0.9400 1.1003 1.3680 1.6300 
Considering (B.45) and (B.46), the total inductance variation of the model winding due the
inward bucking on the second disk is obtained as 4.29 %.
219
Appendix C. Calculation of Series Capacitance of Inter-shield Winding
Appendix C
Calculation of Series Capacitance of

Intershield Winding
C.1 Introduction
Different kinds of transformer winding have been introduced and employed since many
years ago. Among all, the continuous and interleaved windings are more common. The
Intershield Winding (IW) was introduced later by a US registered patent ‎[151] as a disk
type winding which is employed to increase the series capacitance between conductors of
the winding. This, in turn, will decrease the value of initial voltage distribution coefficient
(α) and hence provide a more uniform initial voltage distribution along transformer
windings. Uniform initial voltage distribution will lead to less stress on upper disks (line
end) of the windings. In IW, the shield turns are made of copper or aluminium and placed
between the winding main conductors at predetermined locations while the shield or
shield turns of each disk are insulated from the main conductors of the winding. Some of
major challenges at design stage and manufacturing issues have been discussed in ‎[100].
Since the determination of the series capacitance in intershield winding is not readily
available, a method is proposed in this Appendix together with the detailed calculations.
C.2 Series Capacitance of Intershield Winding
The configuration of the intershield disk winding and its connections were described in
Chapter 3. The equivalent RLC network shown in Fig. C.1 is the ladder network which is
used for transient studies. The total energy for a pair of disks is calculated from (C.1):
Etotal  Ed  Et (C.1)
220
Figure C.1. Equivalent RLC network for the intershield disk winding.
where Ed represents the total disk-to-disk energy and is extracted from (3.28) and Et
represents the total turn-to-turn energy. The calculation of Et using total disk-to-disk
energy was presented in ‎[152]. Hence, turn-to-turn energy calculation for IW is given by:
1
Et  CsU 2  N sh Ett  shield  N sh Ett  shield  2( N  N sh 1) Ett (C.2)
2 up down
where U is the voltage across the pair disks, Nsh is the number of shield turns in each disk.
Ett-shield-up and Ett-shield-down are the energies stored in the upper and lower shield turns which
can be calculated as below:
 2
 
1  2   U  
Ett shield  Ctt  U Vx U Vx    (C.3)
up 2    2 N   
 
 2 2
  U    U  
Ett shield  1 Ctt U Vx 2N    U Vx  2N 1   (C.4)
down 2   2N     2 N   
 
2
 U 
Ett  1 Ctt  (C.5)
2  2 N 
Therefore, substituting (C.3) through (C.5) into (C.2), Et becomes:
221
 2
 
2
 
N 2   U    U 
Et  sh Ctt  U V U V    
x  2 N   
N N 1 Ctt  2 N 
2 x 
sh
 
  (C.6)
 2 2
N   U    U  
 C U V  2 N    U Vx  2 N 1 2 N   
sh
2 tt  x  2N      
 
Appropriate relation between stored energy and the number of shield turns is extracted
using (C.6). Vx is the potential of connected shields between pair disks. In fact, it is an
unknown time-varying quantity over the initial, intermediate or final voltage distribution
on transformer winding. Since the series capacitance of IW is to be found, the initial
impulse voltage distribution of winding has to be taken into account. To figure out this
value, the only solution was to manufacture IW along with other winding types and
perform a practical test. Therefore four windings; one interleaved, one continuous, one IW
with one shield turn and one IW with six shield turns in each disk were manufactured (see
Fig. C.2). A test setup was developed to carry out desired tests on these windings, and a
method was proposed to find the total series capacitance of IW. The proposed method
will pave the way to calculate Vx and eventually calculate the total series capacitance of
transformer winding.
Figure C.2. Manufactured windings from top to bottom: Interleaved, Continuous, Intershield with one shield
turn, and Intershield with six shield turns in each disk.
222
C.2.1 Method to Calculate Total Series Capacitance
The proposed method is completed in two major steps involving theoretical and practical
exercises. Firstly, the voltage between any two neighbor shields (i.e. one shield belongs to
one disk and the other shield belongs to the disk underneath) is measured by means of a 4
GS/s digital oscilloscope. Secondly, according to the winding type, a corresponding
formula is utilized to calculate the total series capacitance in which Vx is a variable (see Fig.
C.3).
223
Input-lead S.P =0
Air-Core
48 4
7 1 3
6 2 2
5 3 4
4 4 3
3 5 2
2 6 1
1
Vx U
49 4
10 3
11 2
12 4
13 3
14 2
15 1
16
S.P =2
4
24 4
23 3
22 2
21 4
20 3
19 2
18 1
17
4
25 4
26 3
27 2
28 4
29 3
30 2
31 1
32
S.P =4
4
40 4
39 3
38 2
37 4
36 3
35 2
34 1
33
4
41 4
42 3
43 2
44 4
45 3
46 2
47 1
48
S.P =6
4
56 4
55 3
54 2
53 4
52 3
51 2
50 1
49
4
57 4
58 3
59 2
60 4
61 3
62 2
63 1
64
S.P =8
4
72 4
71 3
70 2
69 4
68 3
67 2
66 1
65
4
73 4
74 3
75 2
76 4
77 3
78 2
79 1
80 S.P =10
R
Figure C.3. Overall scheme and the sample point of voltage for the simulated winding, (S.P=Sample Point).
The measured Vx is substituted into the formula to determine the total turn-to-turn
capacitance. Calculated turn-to-turn capacitances for the intershield windings are
compared to the windings without shield. In order to validate the calculation results, a
224
standard lightning impulse voltage of reduced amplitude is applied on the manufactured

windings and then initial impulse voltage distribution is measured and compared.
C.2.2 Application of the Proposed Method
Standard reduced impulse voltage is applied on the manufactured windings and the initial
impulse voltage distribution for each winding is measured. In addition, Vx is measured
using a 4 GS/s digital oscilloscope and then (C.6) is used to determine total turn-to-turn
energy. Table C.1 gives the initial voltage distribution percentage in input-lead of the
winding as well as on even disks. According to Table C.1, amount of stress on uppermost
two disks of the continuous disk winding, intershield winding with one shield turn in each
disk, interleaved winding and intershield winding with six shield turns in each disk are
56.4%, 41.8%, 25.8% and 23.8%, respectively.
Table C.1. Initial voltage distribution (%).
Winding Intershield winding

with Intershield winding
Interleaved Continuous
with six shield
winding winding one shield turn in
Sample Point no. turns in each disk
each disk
Input lead 100 100 100 100
S.P 2 74.2 43.6 58.2 76.2
S.P 4 52.7 21.2 34.4 55.1
S.P 6 34 10.1 19.3 35.8
S.P 8 17 4.4 9 17.7
S.P 10 ∼0 ∼0 ∼0 ∼0
Vx -- 77.1 -- 87.4
According to Table C.1, using the intershield winding with six shield turns in each disk will
lead to the most linear initial voltage distribution amongst the four manufactured
windings. Practical measurements revealed that Vx would be around U/2. The measured Vx
is substituted into (C.6) and the total turn-to-turn capacitance for the intershield disk
winding is calculated using (C.7). Equation (C.8) shows that the total turn-to-turn
capacitance of the new intershield disk winding depends on the number of shield turns.
The first part of (C.8) is the same as (3.25) whereas the second part demonstrates the
effect of shields.
1 1  N 1 2 N ( N 1) 
Et  C U 2  C U 2   sh  (C.7)
2 t 2 tt  N2 
 N 
1  N 1  1  2 N ( N 1)  1  N 1 2 N sh ( N 1) 
Ct  C    C  sh   C    (C.8)
2 tt  N 2  2 tt  N tt
 2  N2 N 
225
When the number of shield turns increases, the total series capacitance increases and
consequently α decreases. As a result, the stress on upper disks of the intershield winding
with six shield turns in each disk is less than that of the interleaved winding. Also,
according to Table C.1, the initial impulse voltage distribution along upper disks of the
intershield winding with six shield turns in each disk is closer to the final distribution and
so its α is smaller than other windings.
The inter-disk capacitance for the intershield winding is almost obtained as (3.29).
Therefore, the total series capacitance for the pair-disk intershield winding is given by:
1  N  1 2 N sh ( N  1)  Cdd
Cs  Ctt     (C.9)
 N 
2
2 N 3
226
Appendix D. Glassy Model Transformer
Appendix D
Glassy Model Transformer
D.1 Introduction
Theoretical studies must be proved through practical experiment to reach an acceptable level
of technical satisfaction in research. In fact, practical studies besides theoretical concepts can
complete a scientific circle and lead to a valuable research. Hence, to study the frequency
response a model transformer was specifically designed by the author and manufactured for
this research. This model transformer was used as a test object to conduct various
investigations described in Chapters 3, 4, 5, 6 and 7. The test object specifications as well as
engineering drawing are provided in this Appendix.
D.2 Test Object Overview
This test object has been designed based on power transformer’s concepts. Since the tank of
this transformer was fabricated by plexiglass it is called glassy transformer. Industrial paper
insulations as well as pressboards have been employed in its construction. This test object can
be electrically energised, even though, changing the voltage and transferring the power would
not be simple to be achieved as it has an air-core. The HV winding consists of 8 disks including
8 conductor turns per disk. The LV winding has 10 disks with 6 conductor turns per disk. A 12
mm oil canal including one cylindrical pressboard was designed and implemented between
HV and LV windings while the thickness of the pressboard was 2 mm. The paper insulation
thickness of the HV and LV conductors was 0.5 mm from a side, and the distance between the
HV outermost conductor turns and the tank was 2 mm. The walls were manufactured with
plexiglass. Line and neutral leads of the windings were brought out from the tank through
appropriate HV and LV bushings. Bushings and leads are designed and manufactured based on
standard’s insulation withstands. Internal temperature can be raised up to 220 °C as the
plexiglass used is heat-resistant. An aluminium foil can be wrapped over the glass casing to
simulate the metal transformer tank. A drain valve was installed on the test object cap to
enable the oil injection and also taking oil sample.
227
In addition, internal insulation voltage level is 5kV. When it is filled with transformer oil, any
kind of occurrence as well as oil movement or arcing can be easily observed. Moisture
diffusion (desorption) through paper insulation is quite observable during the winding dry-
out process. The glassy transformer is a portable real transformer designed based on
standard. Its total dimension and weight are 40x40x40[cm3] and 20 kg, respectively.
D.3 Winding Photos and Winding Schematic
Different views of the manufactured test object are shown in Fig. D.1. Figure D.2 provides
the technical drawing for HV and LV windings of the test object.
(a) (b)
(c) (d)
Figure D.1. Manufactured glassy transformer, (a) Side view of winding, (b) Side view without oil, (c) Entire view,
(d) Side view with oil.
228
The cut of X-Y
lUl
] ! """"
61
I A \,...., II
f
Ill
I
!l---2,4,1 0
6
1' I -Gi
7 7
! 111111
6 : 111111
en 5
II i 111111
"'
rrnn
0
~
~ 4 I
I
3 I rrnn
TTTTTTTT
2 ! 1111111
1 : ITITIT 2/.10
! '~/0
6
1 I
I
-- - - 1U2
.~I
- -------H-------
"i l
lUl
IU2
Topic: HV Windin8
(a)
229
' '
' '
' '
' ' ' ' '
' ' ' ' '

' '
' '
' ' ' ' '
' ' ' ' '

' '
' '
' ' ' ' '
' '
' ' ' ' '
' '
10
2 Bandage ' 0.3x20 DIN 40600 ' PSP 3055 0.1
C (Ausr. 02) ' 57 ' I8 27 15 9 ' ' '
9 Z shape 1'hickncs:s: Imm
3 Out e r Protective DIN 40600 PSP 3052 O.I
C (Au sf. 02) ' 57 ' 18 27 15 9 ' ' '
8
4 [n ner Protective Z 6hape ' Thickness: Imm DIN 40600 ' PSP 3052 O.I
A (Aus!. 07) 24.5 25 4
7 spacer ' 4)1'25:r24.5
42 DIN 40600 ' PSP 3052 O.I
B (Aus!. 06) ' 103 ' 4 2 6 2 6 ' ' '
f>
6 Rolling p in +x16.3x103 DL~ 40600 ~SP 3 052 0.1
' ' ' ' '
4
2 .Addi.tlonel lll$\dollon The t.h ickne.s s of each side :O.~ mm Crepe paper 0.2
2 Wire
1 '(3x7x60) ' E- Cu 0.2
I
I HV lfinding Cop~t W.tab\• U • ' Papoer 1fcol.l \ • 0.6 114 Tot.ol •~• bt - 3.1 X,
< PA I A B I c I D E I F G I H I I/J K L I H
""'c
~ Port neme Part dimensions Standard. WatetiaJ ~f.~~
"
Topic: HV Winding Table
page: 04 Number of Po.eca: 01
(b)
230
- l Ul
rz:
-
......*I~
-
4.5 1--
1
3 2 1 6
I
I
{ -
5 4
8
1--
_,
- 1 ( 7
4.5 1-- 1--
- ./
1 ,-
I
6
4.5 1-- 1--
--
I
- 1
I 5
4.5 1-- 1--
- 3 ..J ,- 4
I
4.5 1-- I 1--

-'
- 1 ( 3
4.5 1-- I 1--
,- /
- 3 I
2
4.5 1-- I 1--
1 -- 1
k(.;: 3 2 1 6 5 4 ~~
1U2 -
Topic : Backwf)rd
I
(c)
Figure D.2. HV technical drawing, (a) HV technical schematic, (b) Specifications, (c) Backward step drawing.
231
202
The cut. or X-Y
10
--
~
II
"'~
6
5
:X:
6 ~!
il
-------~ ------
l+ - --I:_J ~I L+-- -f-l

il
U l
I
I
I
I
202
2Ul
Topic: LV W'incUng
(a)
232
' ' ' '

' ' '
' '
' ' ' ' ' '
' ' ' ' ' '

' ' '
' '
: :' : : :' :'
' ' ' ' ' '
' ' ' ' ' '

' ' '
' '
' ' ' ' ' '
' ' ' ' ' '

' ' '
- ' ' '
' '
' ' ' ' ' '
9
2 Bandage 0.3x20 DIN 40600 PSP 3055 0.1
'l'hiekntu; ltnm ' ' ' ' ' '
B
9 FUier tx15.5x21.5 DIN 40600 ' PSP 3052 0.1 ' '
A (Ausf. 07) 20.5 25 4
7
S4 spacer 4xZ5x20 .5 DL~ 40600 'psp 3052 0.1 ' '
B (Ausf. 06) 131 4 ' 2 6 2 6 ' ' ' ' '
6
.a Rolling pin 4x1S.3r131 DIN 40600 psP 3 052 0.1
' ' ' ' ' '
4
2 lnsula.t ion The thickness of each side:O.~mmCrcpc popcr 0.2: '
' ' ' ' ' '
2
1 The Wires (3x7x80) E- Cu o.s
' ' ' ' ' '
1
1 LV Winding ' ' '
L
• PA j A B j c i D j E i F i G i Hj l/J jK i L i H
£,
.1>
:z
Part name Part dimensions I Sl..$nd$rd I llaterial r~i'~1
k Q)
Topic: LV Winding Table
I
(b)
233
.----- 2U2
......
..."~
r.f--
3 2 1 6 5 4 rz
1 I
,- 10
4.5 - -'
' r--
f-- 1 ( 9
4.5 - r--
,- ./
f-- 3 8
4.5 - ' ~
-'
'
r-- 1 r 7
4.5 - I ~
./
r-- 3 '' 6
4.5 - !' ~
_,I
f-- 1 ( 5
4.5 - I ~
r-- 3 ./ ,- 4
4.5 -
' ~
_,'
r-- 1
4.5 -
r
I ~
3
,- ./
f-- 3 2
4.5 -
' ~
_, '
1 1
% 3 2 1 6 5 4 !~
2U1 -
Topie: Backward
I
(c)
Figure D.3. LV technical drawing, (a) LV technical schematic, (b) Specifications, (c) Backward step drawing.
234
Appendix E. Dry-out Process of Model Transformer
Appendix E
Dry-out Process of Model

Transformer
E.1 Introduction
In order to conduct precise study on temperature and moisture impacts on the frequency
response spectrum, the model transformer had to be tested for high and low moisture
contents. High moisture content can be achieved simply through opening the model
transformer valve (stopcock), whereas for low moisture study the test object has to be
dried. This appendix provides information on the dry-out process.
E.2 Model Transformer Dry-out
Water ingress into the transformer tank could be due to the transformer breathing, air-bag
destruction or may be caused by conservator leakage. Cellulose decomposition as well as
insulation aging can also generate water inside the transformer. Maximum and minimum
criteria of permissible water content inside the transformer have been discussed for many
years. These criteria, substantially focused on paper insulation, are less than 4 % and more
than 0.5 %, respectively ‎[65]. Although, some researchers believe that the maximum
permissible limit for transformer water content should be taken as 3 %. The moisture
content of more than 3 or 4 % for transformer can result in bubbling, the formation of free
water and an increased risk of dielectric breakdown, while hyper dry-out and obtaining
less than 0.5 % water in paper insulation may lead to mechanical damage of cellulose
fibres. Therefore, the transformer moisture content should remain within the
predetermined criteria.
Transformer paper insulation has significant tendency to absorb the water specifically in
low temperatures. In fact, a large percentage of the total water inside the transformer is
absorbed through the paper insulation rather than oil insulation. This water should be
removed by the dry-out process to achieve desirable moisture content for transformer.
235
Hence, knowledge of transformer dry-out process is well established and different

methods have been introduced over the years. Low Frequency Heating (LFH), hot air, oil
spray, heat and vacuum with and without cold trap, oil circulation through an oil
circulator, and using an electric oven are among the available methods. Each and every
method has its own advantages. Transformer insulation conditions as well as transformer
size are considered as the major factors to make correct decision on using a specific
method for drying.
In this study, since the paper insulation of the model transformer was completely new, the
application of heat and vacuum without oil was selected as an appropriate method for the
dry-out process. This method requires that insulation oil is dried separately and then
injected into the transformer tank (glassy container). Hence, independently, transformer
oil dry-out was performed through the oil circulation method. To implement both
methods, different equipment had to be provided beforehand and ready prior to starting
the dry-out process. These include oil circulator, appropriate vacuum
pump/pipes/fittings, silica gel, oven for heating the test object, oven for drying silica gel,
three-way stopcock, etc. Figure E.1 shows photos of the various items for the setup.
E.3 Procedure
At first, the vacuum pump was switched on to suck oil from the oil drum into the glassy
anti-vacuum container of the oil circulator. Then, the tap between the oil drum and oil
circulator was closed off to stop the oil flowing towards oil circulator when the desired oil
volume in the glassy container was reached. The oil dry-out station (oil circulator) was
then used to dry the oil. The motor pump of the oil circulator was switched on to circulate
the oil in the circulator; in the meantime the vacuum pump was also running. The oil
heater was also switched on to heat the oil up to maximum 70° C. Heating the mineral oil
more than 70° or 75° C in direct heating systems may gradually burn the mineral oil. The
dry-out process performs poorly when operated at temperatures lower than 40° C ‎[153].
Therefore, during the entire oil treatment, the oil temperature inside the circulator should
be under control through a thermostat. In this study the cold trap was not used as the
mineral oil was new, but the oil circulator was operated continuously for a week. Finally,
an oil sample at 70° C indicated 3 ppm of moisture content in oil. This in turn means that
the mineral oil was dried properly.
Synchronous to oil dry-out, the laboratory oven was utilized to dry-out the test object. The
test object was heated up to 90° C and vacuumed to 750 mTorr for 48 hours to make sure
that moisture is removed from the paper insulation. After that, the metal stopcock on the
236
test object cap was closed to keep the model transformer under vacuum. Then, the output
tap of the oil circulator was connected to the stopcock of the model transformer. A three-
way valve which was connected to a silica gel container was used to break the oil
circulator vacuum. When vacuum was broken through the three-way valve, the metal
stopcock was opened to suck the oil into the model transformer. Finally, the stopcock was
closed and test object was left standing to reach equilibrium between oil and paper
insulations.
(a) (b) (c)
(d) (e) (f)
(g) (h) (i)
237
(j) (k) (l)
(m) (n) (o)
(p) (q)
Figure E.1. Dry-out equipment, (a) Wet silica gel, (b) Half dried silica gel, (c) Silica gel dry-out process (oven
view), (d) Dried silica gel (oven view), (e) Dried silica gel, (f) Silica gel container to break the vacuum after oil
dry-out process, (g) Oil circulator, (h) Internal piping of oil circulator, (i) Oil circulator vacuum gauge, (j)
Transformer oil, (k) Vacuum pump, (l) Internal motor pump of circulator, (m) Glassy anti-vacuum container of
oil circulator, (n) Karl-Fischer equipment, (o) Metal stopcock, (p) Vacuum gauge for test object dry-out, (q)
Model transformer dry-out process (oven view).
238
Appendix F. Moisture Content Recognition in Transformer
Appendix F
Recognition of Moisture Content in

Transformer
F.1 Introduction
Having continuous information about the trend of change of insulation system condition
and internal mechanical integrity is vitally important for the system operator and asset
owner. With uninterrupted flow of information, management of effective life span,
scheduling maintenance operations for transformers can be promoted.
Existence of humidity in the insulation systems containing oil-impregnated paper may

result in premature aging that may lead to abrupt temperature changes. This may in turn
result in the production of bubbles and internal partial discharges within transformers.
Naturally, normal aging of cellulosic material produces water. This is caused by the
molecular breakdown of the cellulose which is high in hydrogen and oxygen molecules.
The original bonds are broken by what is called de-polymerization process and loose
hydrogen and oxygen molecules reunite to form H2O, hence water ‎[154].
In addition, exposure during maintenance or repairs is another avenue for water to find its
way into the transformer. Consequently, for the evaluation of the insulation system
viability, the extent of humidity in the impregnated paper is required. In this regard, the
only direct method to determine the moisture content in the cellulose is by testing paper
samples taken from transformer. However, there are several indirect methods to
determine the moisture content as well.
A popular indirect method for determining humidity in a transformer is using oil sample.
Oil sample can be easily taken from transformer while it is service. This sample is then
analysed by what is called Karl Fisher Titration (KFT) test.
Karl-Fischer Titration as a non-destructive test with the interpolation capability from the
equilibrium curve of oil-impregnated paper is currently common in many countries and
239
transformer operators use it continuously. In view of the researches conducted, the

aforementioned method is deemed as a conventional method for the measurement of the
humidity of the insulation paper of the transformer under operational conditions ‎[155],
which will be duly discussed in this Appendix. Another indirect method is capacitance
sensor method which will be discussed in detail in Subsection F.2.2.
The method of Polarization and Depolarization Current (PDC) is regarded as a non-

destructive and simple method for measurement of dielectric response function, as well as
conductivity and humidity rate for oil-paper insulation in a transformer ‎[139]. Other
methods of significance are Recovery Voltage Measurement (RVM) and Frequency Domain
Spectroscopy (FDS). In this Appendix, these various approaches for humidity evaluation in
oil immersed transformers will be discussed.
F.2 Water Content Recognition in Oil-impregnated Paper
F.2.1 Karl-Fischer Titration
In order to measure the moisture content of the paper, oil is sampled from an energized
transformer. Then, KFT instrument is employed to provide the moisture content of the
sampled oil. The last step is to utilize equilibrium curves to determine the water content of
the oil-impregnated paper in percentage. KFT has several considerations, some of which
are mentioned hereunder:
1. Errors resulting from unsuitable and non-standard sampling;

2. Temperature changes in the oil sample, at the time of sampling until delivery of the
sample to the laboratory for the due testing;
3. Unsuitable storage;
4. Different sample preparation in laboratories. In addition, there are various
available titration systems that use different techniques;
5. Existing errors in the equilibrium curves for low humidity cases and also low
temperature analysis.
These considerations should be taken into account to reach to maximum accuracy. For
instance, sampling errors is one of the problems caused by samplers. Experience shows
that ambient temperature and humidity has direct impact on the test results. Hence,
samples for testing taken in a damp environment have significant difference with samples
taken in a dry environment. Therefore, it is always recommended to take oil samples
using special syringes ‎[156], so as to minimize the relevant error. Another problem with
oil sampling is the temperature variation of the test sample from the sampling location to
240
the lab. Also, samples shall at no time be exposed to direct sunlight. In addition, the time
between collection and analysis of the sampled oil should be not exceeding seven days
‎[156]. If the samples are left for a period more than seven days in the laboratory,
variations of the temperature will adversely impact the sample which in turn affects test
results. Obviously, when the experiment including sampling and KFT is run in the
laboratory as to what conducted in this study, there is no concern on this issue.
F.2.1.1 KFT Oil Sampling
In order to measure the water content of the transformer paper insulation through the
transformer oil, the oil sample should be taken from the transformer tank/container. The
sample can be taken using laboratory glass bottle (250 or 500 ml) or through glass
syringe. To avoid the penetration of environment humidity into the sample, it is strictly
recommended that oil sample is taken by the syringe having plastic stopcock rather than
laboratory glass bottle. Figure F.1 shows the laboratory glass bottle and glass syringe.
The plastic stopcock, snapped to the syringe, must remain firmly affixed to it at all times to
prevent leaks and tightly close the syringe for transportation to laboratory ‎[157]. The
sample should not be subjected to vacuum to avoid any moisture penetration. Then, the
sample should be shipped to laboratory and KFT is performed. Taking oil sample for KFT
is not considered to be as accurate as DGA (Dissolved Gas Analysis), but it has its own
concerns. In this study, the oil sample at different temperatures from the model
transformer was taken by a glass syringe, and then tested by KFT immediately (< 5 min) to
get maximum accuracy. Figure F.2 illustrates an oil sample taken from the test object.
(a) (b)
Figure F.1. Oil sample containers, (a) Laboratory glass bottle, (b) Glass syringe.
241
(a) (b)
Figure F.2. Oil sample containers, (a) Side view, (b) Front view.
This volume of oil sample (20 ml) is enough for KFT measurement. At least, 10
measurements on a 20 ml sample can be conducted and the average of results is taken as
the measured value.
F.2.1.2 KFT Titration
Karl Fischer (1901-1958) was a chemist working at a petrochemical company in Germany

in the 1930’s. He published a method to determine trace amount of water samples. This
method was a technique to recognize the moisture content. Nowadays, Karl Fischer
method is an analytical technique used worldwide to measure the moisture (water)
content in solids, liquids or gases.
Titration is defined as “A technique to determine the concentration of a substance in

solution by adding to it a standard reagent of known concentration in carefully measured
amounts until a reaction of definite and known proportion is completed, as shown by a
colour change or by electrical measurement, and then calculating the unknown
concentration” ‎[158].
K-F titration involves two reactions. In the first reaction, an alcohol (usually methanol or
ethanol), sulphur dioxide (SO2) and a base (RN) react to form an alkyl sulphite
intermediate ‎[158]:
CH3OH  SO2  RN  [RNH ]SO3CH3 (F.1)
In the second reaction, the alkyl sulphite reacts with iodine (I2) and the water from the
sample:
242
[RNH ]SO3CH3  I  H O  2RN  [RNH ]SO4CH3  2[RNH ]I (F.2)

2 2
Since water and I2 are consumed in equimolar amounts in reaction F.2, if the amount of I2
consumed is known, the amount of water that was present in the sample can be
determined.
F.2.1.3 KFT Equipment
Based on the method introduced by Karl Fischer, KFT equipment has been developed and
commercialised. Figure F.3 shows an example of KFT equipment together with accessory
(micro scale).
Two types of KFT are commercially available:
1. Volumetric, and
2. Coulometric.
In fact, both methods use bio potentiometric titration to find the amount of I2 consumed by
the water. Electrical conductivity variation of the reaction solution is the thing called bio
potentiometric titration. This technique is pretty similar to direct titration; an indicator
and reference electrodes are used and overall electric potential is calculated to reach
desirable titrant.
Indeed, coulometric which is used in this study aimed to explore the amount of matter
transformed during electrolysis reaction by measuring the produced or consumed
coulombs. Based on this, KFT equipment using coulometric contains various components
as illustrated in Fig. F.4. Detail information of the equipment is provided in ‎[159].
(a) (b)
Figure F.3. KFT equipment, (a) Digital micro scale, (b) Glassy container, main unit and keypad.
243
(a) (b) (c)
(d) (e) (f)
(g) (h) (i)
Figure F.4. KFT unit main components, (a) Double platinum wire electrode (0.8 x 4 mm), (b)Generator
electrode for Karl Fischer titrations, with diaphragm, (c) KF absorber tube for coulometer cell, (d) SGJ stopper,
(e) Stopper, (f) Plastic tube, (g) Titration vessel holder for coulometric cells, (h) KF titration vessel, (i) Keypad
for 756 KF Coulometer.
F.2.1.4 KFT Equilibrium Curves
In order to determine the paper humidity through oil humidity, oil-paper equilibrium
curves have been derived in 1960 for the first time as an indirect method to measure
paper humidity by Fabre and Pichon ‎[160], ‎[110]. After that Fallou worked on Fabre-
Pichon curve and summarized the moisture content of the oil and oil-impregnated
pressboard by KFT reaction method ‎[110]. In 1963, Norris ‎[162] did further work on this
and later published equilibrium curves known as “Norris curves”. Twenty years after that,
Oommen ‎[161] developed a set of moisture equilibrium curves ‎[161] as shown in Fig. F.5.
It is stated in ‎[110] that Oommen used and combined moisture in oil versus relative
humidity curves in air and moisture in paper versus relative humidity curves in air to
244
reach moisture in oil versus moisture in paper equilibrium curves. In ‎[110], it has been
also discussed that Griffin et al ‎[163] derived similar curves for mineral oil and paper in
1988 using Oommen’s method. In 1999, a careful study by Du et al ‎[110] provided
discussion on all available equilibrium curves. Since, most studies provided the oil-paper
equilibrium curves up to at most 100 ppm, they have used the Oommen’s method and
tried to extract a wider range of moisture concentration in oil insulation. This work was
conducted from 0° C to 100° C and oil-paper equilibrium curves extracted up to 800 ppm
(see Fig. F.6.). Indeed, this attempt was crucial for high moisture concentration in oil-
paper systems. Therefore, due to the wide range of humidity in oil-paper system, the MIT
equilibrium curves were employed in the current study to derive the moisture content.
This in turn provided opportunity to figure out the moisture content in oil insulation of
glassy model transformer for a wide range, and then to derive paper water content.
245
Figure F.5. Oommen equilibrium curves for oil-paper system ‎[161].
Figure F.6. MIT-developed curves for water equilibrium in cellulose/mineral oil systems ‎[110].
246
F.2.2 Capacitor Method
Using capacitive probes is another method to measure moisture in oil. Capacitive probes
measure the changes of capacitance caused by water molecules. Water molecules migrate
into the dielectric of the capacitor and change the capacitance which is measurable.
Indeed, the thin-film polymer as shown in Fig. F.7 absorbs or releases water vapour as the
relative humidity of the ambient air rises or drops ‎[164]. Electrical properties of the
polymer film depend on the amount of water contained in it. The associated instrument
has a sensor that is inserted into the oil sample and measures the capacitance of the
sensor. This capacity is converted into an oil humidity reading. Finally, equilibrium curve
is used to recognize impregnated paper humidity.
Figure F.7. Capacitance sensor layers.
F.2.3 Paper Sample Method
The Dean and Stark (Deanastarka) procedure can be employed to measure the water
content of a diverse range of samples. This method has been used in industrial
laboratories to measure water in petroleum oils and can also be performed to determine
water content in transformer paper.
As an advantage, water content percentage of transformer paper insulation can be directly

recognized through this technique. The sample of paper insulation in transformer is mixed
with a solvent (usually a toluene/xylene mix) and refluxed under a condenser using a
special receiver that is illustrated in Fig. F.8.
There are two common designs of receivers, one for solvents that are heavier than water,
and the more common one for solvents that are lighter than water. The water and solvent
are refluxed, and as they condense the two phases separate as they run into the receiver.
247
(a) (b) (c) (d)
Figure F.8. Dean-Stark apparatus, (a) Condenser, (b) Receiver, (c)Flask, (d) Assembled parts ‎[165].
Practically, a sample from the transformer paper insulation is taken. Then, this sample is
cut into maximum 10×10 cm scraps. The flask is filled up to 2/3 with solvent and scraped
papers are added to the solvent afterwards. The flask is heated up for 2 hours to condense
water vapour and finally water drops are gathered in the receiver. To calculate the
humidity in transformer paper insulation; the moisture content is given by
 
Moisture(%)   Y 100 (F.3)
 PK 
 m
where Y represents mass of accumulated water, P is mass of paper scraps and Km

represents mass coefficient of oil. Generally, Km=0.7 and Km=1 are taken for impregnated
and un-impregnated oil paper, respectively. Figure F.9 shows the practical setup of Dean-
Stark method.
(a) (b)
Figure F.9. Different views of measuring water content of the sampled papers using Dean-Stark method.
248
F.2.4 Electrical Methods
In order to facilitate the measuring of paper humidity without using oil samples, three
electrical methods are commonly considered worldwide. They are Polarization and
Depolarization Current (PDC), Frequency Domain Spectroscopy (FDS), and Recovery
Voltage Measurement (RVM) ‎[51].
F.2.4.1 Polarization and Depolarization Current (PDC)
This method is based on the fact that when a DC voltage is applied across a sample of
insulation material, the interaction of the electrical field with the free charges and other
types of limited charges results in the movement of electrical charges inside the insulation
material and an electric current is created, which is known as polarization. This current
over time decreases gradually and reaches a fixed value, which is related to the
conductivity of the insulation material. If the supply is removed when the polarization
current reaches to the fixed value and the two ends of the insulation material are short-
circuited, the depolarization current leaps towards a negative value and then decreases to
zero. If the polarization time is sufficient, the depolarization current shall be suitable for
the function of the dielectric reaction.
The polarization and depolarisation currents can be measured precisely by a micro-

ampere meter. Generally, function of DC charge can be employed to characterize the
influence of material properties, such as conductivity, on the dielectric response of the
insulation material. This method is used to evaluate the moisture content of cellulosic
material in power equipment.
F.2.4.2 Recovery Voltage Measurement (RVM)
RVM is based on established knowledge on the phenomenon of the polarization of oil-

impregnated paper insulation. There are different types of polarization. In the case of
moist oil/paper insulation, there is a polarization due to the water molecules contained in
the insulation material. By applying a DC voltage, these molecules, which are electrically
neutral, acquire a polarity and try to drift in the direction of the electrical field. That means
that molecules get energized. The applied voltage is then removed, the terminals are
short-circuited to allow discharging and opened afterwards. Some energy is still stored in
the molecules. A voltage due to this stored energy can be measured, which is called the
recovery voltage. Since different recovery voltages are obtained with the change of time
duration, these voltages shall be different for various insulation materials with different
humidity. By this method, the insulation condition is examined by tracing the polarization
249
spectrum. However, some research works have revealed excessive dependency of the
method on the geometry, temperature and conductivity parameters of oil. Therefore, RVM
is regarded as an outdated method nowadays ‎[166].
F.2.4.3 Frequency Domain Spectroscopy (FDS)
Instead of studying the polarization process in time domain, the study of dielectric
response could be conducted in frequency domain when an AC sinusoidal voltage U(ω) is
applied ‎[167]. The complex impedance and dielectric dissipation factor (tan δ) are
measured. In this method, the imaginary and real parts of the capacity of an insulation
system are estimated over a frequency range and humidity amount corresponding to this
capacity is determined using diagram comparison method.
If the effective relative dielectric susceptibility is defined as (7.9) and complex permittivity
is given by (7.10); then the dielectric dissipation (loss) factor shall be estimated as (F.4):
   
  ( )    
tan  ( )   0 
(F.4)
   ( ) 
 
 r

 
where, all the parameters have been defined in Chapter 7. Figure F.10 shows variation of
loss factor of oil-immersed paper with frequency.
Figure F.10. Variation of loss factor of oil-immersed paper with frequency and dominant influences ‎[119],
taken and modified.
250
Appendix G. Study on Recommended Solution
Appendix G
Study on Recommended Solution
G.1 Introduction
In Chapter 7, a technique to distinguish the insulation characteristic impacts from the

mechanical deformation influence on the FRA trace was recommended. The procedure to
conduct this technique was given through a chart, and the criteria highlighted. This
Appendix provides a numerical example on this recommended solution.
Before performing evaluation through the statistical indicators, reference and measured
FRA traces should be examined through the recommended technique to determine the
next appropriate action. This Appendix is also provided the results of winding
deformation versus the temperature and moisture influences on model transformer for
comparison.
G.2 Case Study 1
In order to test the recommended method, FRA spectra of HV winding of the glassy model
transformer taken at 30 and 90 °C were examined. A program was developed, and
according to the procedure presented in Fig. 7.15, FRA spectra (Xi and Yi) were loaded in
the program. Afterwards, î and î were calculated. We need to calculate î and î to
derive the resonant frequencies. In fact, wherever î and î experience zero value a
resonant peak has happened. Figure G.1 shows the references (Xi and Yi) and derivative
spectra ( î and î) for the HV winding of the test object at 30 and 90 °C.
Z{ î }n and Z{ î }m should then be calculated. This in turn helps to realize number of

resonant points and ascertain whether the winding is mechanically deformed. n≠m will
require statistical indicators to be utilized as an evaluation method, while n=m reveals
normal mechanical condition for the winding. Therefore as the next step, S{ î } and S{ î }
were calculated and Z{ î }n and Z{ î }m were derived. Table G.1 shows the calculated
values for Z{ î }n and Z{ î }m. According to Table G.1, n=m and thus FRA statistical
251
indicators are not required to be considered. Therefore, Rn ≜ R{Xi, Yi}n were calculated and
illustrated in the last column of Table G.1.
80
60 4.267 MHz
40
20
Magnitude [dB]
-20
-40
-60
1.223 MHz
-80
1.861 MHz Derivative spectrum
-100 HV winding spectrum (at 30 °C)
6 7
10 10
Frequency [Hz]
(a)
80
4.178 MHz
60
40
20
Magnitude [dB]
-20
-40
1.802 MHz
-60
1.184 MHz
-80
Derivative spectrum
HV winding spectrum (at 90 °C)
-100 6 7
10 10
Frequency [Hz]
(b)
Figure G.1. Reference and derivative spectra for HV winding of glassy model transformer, (a) Spectra at 30 °C,
(b) Spectra at 90 °C.
252
Table G.1. Calculated values for Z{ î}n at (30°C) and Z{ î}m at (90°C).
Z{ î }n [Hz] Z{ î }m [Hz] Rn ≜ R{Xi, Yi}n

1223110 1184837 0.968708456
1740778 1689525 0.970557417
1861823 1802298 0.968028647
2347992 2277552 0.969999898
2481509 2407063 0.969999706
2565236 2488279 0.970000031
2771751 2688598 0.969999830
4267881 4178285 0.979006912
4559604 4422816 0.970000026
7253854 7038179 0.970267530
7510670 7275650 0.968708517
7927191 7689375 0.969999966
8952855 8684270 0.970000073
9359869 9077133 0.969792740
10686194 10365608 0.969999983
11546489 11200095 0.970000058
11804773 11550630 0.978471166
12213072 11836980 0.969205782
13480431 13176018 0.977418155
15067130 14605416 0.969356208
15393943 14932125 0.970000019
15913342 15535942 0.976284051
18182201 17627035 0.969466513
18578695 18221334 0.980765011
To obtain α and β, the moisture content of the reference trace was used, equal to 4 %, as it
was measured earlier in Chapter 7. Upper limits (Wul) and lower limits (Wll) for moisture
content were taken as 4 % and 0.5 % respectively. Also the WCP change for this case was
taken as 0.5 (see Table 7.8). According to (7.16) and (7.17), α and β are obtained as:
  1 0.79 (0.5  4)  0.9447 (G.1)

0.5 100
(4  4)
  1  0.79 1 (G.2)
0.5 100
Considering calculated values for Rn ≜ R{Xi, Yi}n in Table G.1, all values satisfy the criteria in
(G.3). It means that the winding has “Normal” condition and “No Action Required”. Based
on our knowledge about the test object in Chapter 7, the calculated result in (G.3) is quite
reasonable for this case.
0.9447    R    1 (G.3)
n
253
G.3 Case Study 2
In order to study the winding deformation recognition through the recommended

solution, the glassy model test object was again examined.
At first, the frequency response trace of the model transformer was recorded from the HV
side while the LV side was left open circuit and test object tank did not have oil. Next, to
model a mechanical defect in the test object, the modification involved the short circuit on
LV terminals was performed. This modification can in turn block the flux flow in the
transformer air core and model winding internal short-circuit on LV side. Figure G.2
shows the frequency response spectra for HV winding when the LV winding was left open-
and short-circuited, respectively.
-10
-20
Magnitude [dB]
-30
-40
-50
-60
Measured Trace (HV winding)
Original Trace (HV winding)
-70 4 5 6 7
10 10 10 10
Frequency [Hz]
(a)
254
-10
-20
Magnitude [dB]
-30
-40
-50
-60
Measured Trace (HV winding)
Original Trace (HV winding)
-70 5 6 7
10 10 10
Frequency [Hz]
(b)
Figure G.2. Frequency response spectra for HV winding when the LV winding was left open and short-circuited
(test object without oil), (a) Entire frequency band (20 Hz – 20 MHz), (b) The area enclosed by dash-line
rectangle in Fig. G.2(a), (100 kHz – 20 MHz).
According to Fig. G.2, when the LV winding is short-circuited, changes to the frequency
response of the HV winding occurred in the range from 20 Hz to 3 MHz. For frequencies
above 3 MHz, there is no significant discrepancy between recorded spectra. The reason
lies in the fact that the winding self-inductance has changed due to the short circuit
deliberately created. In addition, in the case of short-circuit, the low frequency band of
FRA spectrum (to the first anti-resonance) shows a different trend as compared to the
original spectrum. In fact, its falling trend is slightly moderate. This means that
transformer HV winding experiences less inductance when having internal short-circuit
than ‘normal’ state in LV winding. The first minimum peak has shifted to higher frequency
and its magnitude is also reduced. Similar result for inductance reduction in transformer
winding was obtained on a different test object in Chapter 6, subsection 6.6.2.
To examine the recommended solution S{ î } and S{ î } were calculated and Z{ î }n and

Z{ î }m were derived. Table G.2 shows the calculated values for Z{ î }n and Z{ î }m.
According to Table G.2, n≠m and the transformer is suspected to have mechanical defect.
Therefore, FRA statistical indicators are required to be taken into consideration. The
reference and measured FRA spectra as well as their derivative curves for HV winding are
illustrated in Fig. G.3.
255
10
0
Magnitude [dB]
-10
-20
1.840 MHz
-30
Derivative spectrum
-40 Refrence spectrum (HV winding)
5 6 7
10 10 10
Frequency [Hz]
(a)
10
0
Magnitude [dB]
-10
-20
-30
2.102 MHz
Derivative spectrum
-40 Measured spectrum (HV winding, short circuit in LV)
5 6 7
10 10 10
Frequency [Hz]
(b)
Figure G.3. HV winding spectra (a) Reference spectrum and its derivative (100 kHz – 20 MHz), (b) Measured
spectrum and its derivative (100 kHz – 20 MHz).
256
Table G.2. Calculated values for reference spectrum Z{ î}n, and measured spectrum Z{ î}m.
Z{ î }n [Hz] Z{ î }m [Hz]
1840822.911 2102123.015
2961946.861 3382387.641
3028202.732 3496511.071
3420008.789 3654687.782
3535401.565 3777998.756
3654687.782 4765873.542
3777998.756 4872481.525
4765873.542 5626005.61
4872481.525 6284034.927
5688581.755 8952855.389
6284034.927 9673607.699
9153122.082 9781203.931
10805053.01 10805053.01
11674917.05 11546489.36
12203072.06 12203072.06
12896990.82 12896990.82
14891496.72 14891496.72
15913342.45 15913342.45
18578695.3 17972300.97
18785339.78 18172200.75
---- 18578695.3
---- 18785339.78
G.4 Winding Deformation vs. Temperature and Moisture Influences on FRA

Spectrum
To compare the frequency bands influenced by winding deformation and

Temperature/Moisture variation in FRA spectrum, some mechanical changes were
emulated on model transformer and the results are compared with insulation parameters
changes.
G.4.1 Influence of Internal Short-circuit
At first, the frequency response trace of the model transformer is recorded from the HV
side while the LV side is left open circuit. In order to study the internal short circuit, the
modification involved the short circuit on LV terminals is performed on the test object.
This modification can in turn block the flux flow in the transformer air core and model
winding internal short-circuit. Figure G.4(a) shows the frequency response spectra when
the LV winding is left open- and short-circuited, respectively. According to Fig. G.4(a),
when the LV winding is short-circuited, changes to the frequency response of the HV
winding occurred in the range from 20 Hz to 3 MHz. For frequencies above 3 MHz, there is
no significant discrepancy between recorded spectra. The reason lies in the fact that the
257
winding self-inductance has changed due to the short circuit deliberately created. In
addition, in the case of short-circuit, the low frequency band of FRA spectrum shows
different trend to reach to the first anti-resonance as compared to the original spectrum.
In fact, its falling trend is slightly moderated. This means that transformer HV winding
experiences less inductance when having internal short-circuit than ‘normal’ state in LV
winding.
G.4.2 Influence of Tank Grounding
In order to conduct this experiment, the aluminium tank of the test object was grounded
through laboratory earth. FRA trace for the HV winding was measured while LV terminals
were left open circuit. HV winding FRA spectra for isolated and grounded tank are shown
in Fig. G.4(b). According to Fig. G.4(b), the number of resonance frequencies for the case of
grounded tank has increased as compared to the un-grounded case, in particular the mid-
frequency band. It can be explained through shunt capacitance increment. In addition, the
magnitude of the first anti-resonance in HV trace has varied considerably due to changes
in conductance between HV and LV windings with respect to the tank. This can be
interpreted by winding loss factor (conductivity). In fact, any changes in winding loss
factors can result in FRA magnitude of resonance/anti-resonance points to change
accordingly. The frequency band 300 kHz – 20 MHz is affected under such a circumstance.
These results are already compared with moisture and temperature influences on FRA
spectrum on similar winding discussed in Chapter 7 for the model test object to highlight
the affected frequency bands due to different incidents.
0
-10
-20
Deviated Frequency Band
Magnitude [dB]
-30
-40
-50
Frequency band: 4 kHz - 20 MHz

-60
HV winding (orginal spectrum)
HV winding (short circuit in LV winding)
-70 4 5 6 7
10 10 10 10
Frequency [Hz]
(a)
258

-10
-20
Magnitude [dB]
-30
-40
-50

-60
HV winding spectrum (Grounded Tank)
HV winding spectrum (Isolated Tank)
-70 4 5 6 7
10 10 10 10
Frequency [Hz]
(b)
0
-10
-20
-30
Magnitude [dB]
-40
-50
-60
-70 Frequency Band: 5 kHz - 20 MHz

4 5 6 7
10 10 10 10
Frequency [Hz]
(c)
Figure G.4. (a) HV winding frequency response spectrum when LV winding is open-circuited (original
spectrum) and short-circuited (affected frequency-band 20 Hz-3 MHz),(b) FRA spectra of HV winding for
isolated and grounded tank (affected frequency-band 300 kHz-20 MHz), (c) FRA spectra for HV winding due to
moisture migration from paper into the oil insulation at 30°C and 90°C, replotted from Fig. 7.4 for comparison
(affected frequency-band 800 kHz-20 MHz).
Based on comparison, it can be summarized that the influence of moisture migration

appears in the mid- and high-frequency band of the FRA spectrum, the indicated frequency
values are estimated in Fig. G.4 caption. In contrast, the mechanical defects can influence
259
each and every part of the entire frequency band randomly. Indeed, FRA trace deviation
due to mechanical defects is certainly influenced by the particular type of deformation.
Self- and mutual inductances, series and shunt capacitances or even winding resistance
variation due to the winding deformation has its own impacts on the FRA trace. As it was
discussed in previous section, for accurate interpretation of the FRA results, these impacts
should be distinguished from the moisture influence using recommended thechnique.
G.5 Conclusion
The recommended solution to distinguish the insulation characteristic impacts from the
mechanical deformation influence on the FRA trace was examined on deviated FRA
spectra. It led to a correct decision on the test object. This solution is recommended to be
implemented as preliminary stage in FRA interpretation.
260
Bibliography
Bibliography
[1] IEC Standard 60076-18; Measurement of frequency response, Edition 1.0, 2012-07.
[2] IEEE Std. C57.149; IEEE guide for the application and interpretation of frequency response
analysis for oil-immersed transformers, 2012.
[3] E. P. Dick and C. C. Erven, “Transformer diagnostic testing by frequency response analysis”,
IEEE Trans. Power App. Syst., Vol. PAS-97, No. 6, pp. 2144–2153, 1978.
[4] M. Florkowski, J. Furgal and P. Pajak, “Analysis of fast transient voltage distributions in
transformer windings under different insulation conditions”, IEEE Trans. Dielectr. Electr.
Insul., Vol. 19, No. 6, pp. 1991-1998, 2012.
[5] M. Florkowski and J. Furgał, “Application of transfer function to recognition of resonance

overvoltages in transformer windings”, Meas. Sci. and Technology, Institute of Physics
Publishing, Vol. 21, No. 12, pp.1-9, 2010.
[6] K. Ludwikowski, K. Siodla and W. Ziomek, “Investigation of transformer model winding

deformation using sweep frequency response analysis”, IEEE Trans. Dielectr. Electr. Insul.,
Vol. 19, No. 6, pp. 1957-1961, 2012.
[7] C. Q. Su, Electromagnetic Transients in Transformer and Rotating Machines Winding, IGI
Global press, 2012.
[8] M. Bagheri, Mohammad S. Naderi, T. Blackburn and B.T. Phung, “Bushing characteristic
impacts on on-line frequency response analysis of transformer winding”, IEEE Int’l. Conf.
Power Energy (PECON), Kota Kinabalu, Malaysia, pp. 921-926, 2012.
[9] M. Bagheri, Mohammad S. Naderi, T. Blackburn, B.T.Phung and Z.Liu “Frequency response
analysis to recognize inductance variation in transformer due to internal short circuit”, IEEE
Int’l. Power Energy Conf. (IPEC), Ho Chi Minh City, Vietnam, pp. 677-681, 2012.
[10] M. Bagheri, Mohammad S. Naderi, T. Blackburn, “Advanced transformer winding deformation

diagnosis: moving from off-line to on-line”, IEEE Trans. Dielectr. Electr. Insul., Vol. 19, No. 6,
pp.1860-1870, 2012.
[11] M. Florkowski and J. Furgal, “Transformer winding defects identification based on a high
frequency method”, Meas. Sci. and Technology, Institute of Physics Publishing, Vol. 18, No. 9,
pp. 2827-2835, 2007.
[12] E. Rahimpour, M. Jabbari and S. Tenbohlen, “Mathematical comparison methods to assess

transfer functions of transformers to detect different types of mechanical faults,” IEEE Trans.
Power Del., Vol. 25, No. 4, pp. 2544-2555, 2010.
[13] P. Karimifard, G. B. Gharehpetian and S. Tenbohlen, “Localization of winding radial

deformation and determination of deformation extent using vector fitting-based estimated
transfer function”, European Trans. Electr. Power, Vol. 19, No. 5, pp. 749-762, 2009.
261
Bibliography
[14] P. Mukherjee and L. Satish, “Construction of equivalent circuit of a single and isolated
transformer winding from FRA data using the ABC algorithm,” IEEE Trans. Power Del., Vol.
27, No. 2, pp. 963-970, 2012.
[15] M. Bagheri, Mohammad S. Naderi, T. Blackburn and T. Phung, “FRA vs. short circuit
impedance measurement in detection of mechanical defects within large power
transformer”, IEEE Int’l. Sympos. Electr. Insul. (ISEI), Puerto Rico, USA, pp. 301-305, 2012.
[16] K. G. N. B. Abeywickrama, Y. V. Serdyuk and S. M. Gubanski, “Exploring possibilities for

characterization of power transformer insulation by frequency response analysis (FRA)”,
IEEE Trans. Power Del., Vol. 21, No. 3, pp. 1375-1382, 2006.
[17] M. Bagheri, Mohammad S. Naderi, T. Blackburn and B.T.Phung, “Frequency Response

Analysis vs. Flux Division Measurement in detection of transformer winding internal short
circuit”, IEEE Int’l. Conf. Power Sys. Tech. (Powercon), Auckland, New Zealand, pp. 1-5, 2012.
[18] M. Bagheri, Mohammad S. Naderi, T. Blackburn and B.T.Phung, “Frequency response analysis
and short circuit impedance measurement measurement in detection of winding
deformation within power transformers”, IEEE Electr. Insul. Mag.,Vol. 29, No. 3, pp. 33-40,
2013.
[19] A. A. Reykherdt, V. G. Davydov, “Case studies of factors influencing frequency response

analysis measurements and power transformer diagnostics,” IEEE Electr. Insul. Mag., Vol. 27,
No. 1, pp. 22-30, 2011.
[20] J. R. Secue and E. Mombello, “Sweep frequency response analysis (SFRA) for the assessment
of winding displacements and deformation in power transformers”, Electric Power Systems
Research, Vol. 78, pp. 1119-1128, 2008.
[21] P.M. Nirgude, D.Ashokraju, A.D. Rajkumar and B.P.Singh “Application of numerical eveluation
techniques for interpreting frequency response measurements in power transformers”, IET
Science, Measurement and Technology, Vol. 2, No. 5, pp. 275-285, 2008.
[22] T. Y. Ji, W. H. Tang and Q. H. Wu, “Detection of power transformer winding deformation and
variation of measurement connections using a hybrid winding model”, Electric Power
Systems Research, Vol. 87, pp. 39-46, 2012.
[23] S. A. Ryder, “Methods for comparing frequency response analysis measurements,” IEEE Int’l.
Sympos. Electr. Insul., Boston, MA, USA, pp. 187-190, 2002.
[24] J. W. Kim, B. Park, S.C. Jeong, S. W. Kim and P. G. Park “Fault diagnosis of a power transformer
using an improved frequency-response analysis”, IEEE Trans. Power Del., Vol. 20, No. 1, pp.
169-178, 2005.
[25] A.Kraetge, M. Kruger and P. Fong, “Frequency response analysis- status of the worldwide
standardization activities”, Int’l Conf. Condition Monitioring and Diagnosis, Bejing, China, pp.
651-654, 2008.
[26] G. M. Kennedy, A. J. McGrail and J. A. Lapworth, “Using Cross-Correlation coefficients to

analyze transformer Sweep Frequency Response Analysis (SFRA) traces”, IEEE Power Eng.
Soc. Conf. and Expos. in Africa (PowerAfrica), pp. 1-6, 2007.
[27] Z. Wang, J. Li and D.M. Sofian, “Interpretation of Transformer FRA responses – Part I:
Influence of Winding Structure”, IEEE Trans. Power Del., Vol. 24, No. 2, pp.703-710, 2009.
[28] S. A. Ryder, “Diagnosing transformer faults using frequency response analysis”, IEEE Electr.
Insul. Mag., Vol. 19, No. 2, pp. 16-22, 2003.
262
Bibliography
[29] D. M. Sofian, Z. Wang and J. Lie, “Interpretation of Transformer FRA Responses- Part II:
Influence of Transformer Structure”, IEEE Trans. Power Del., Vol. 25, No. 4, pp. 2582-2589,
2010.
[30] T. De Rybel, A. Singh, J. A. Vandermaar, M. Wang, J. R. Marti and K. D. Srivastava, “Apparatus

for online power transformer winding monitoring using bushing tap injection”, IEEE Trans.
Power Del., Vol. 24, No. 3, pp. 996-1003, 2009.
[31] M. Bagheri, Mohammad S. Naderi, T. R. Blackburn and B.T. Phung, “Practical challenges in
online transformer winding deformation diagnostics”, Int’l. Conf. Electric Power and Energy
Conversion Syst. (EPECS), Sharjah, UAE, pp. 1-6, 2011.
[32] V. Behjat, A. Vahedi, A. Setayeshmehr, H. Borsi and E. Gockenbach, “Diagnosing shorted turns
on the windings of power transformers based upon online FRA using capacitive and
inductive couplings”, IEEE Trans. Power Del., Vol. 26, No. 4, pp. 2123-2133, 2011.
[33] M. Bagheri, Mohammad S. Naderi, T. Blackburn, T. Phung and H. Haeri, “On-line transformer
winding deformation diagnosis: A profound insight to methods,” 26th Int’l. Power Syst. Conf.
(PSC), Tehran, Iran, pp. 1-14, 2011.
[34] A. Setayeshmehr, H. Borsi, E. Gockenbach and I. Fofana, “Online monitoring of transformer

via transfer function”, IEEE Electr. Insul. Conf. (EIC), pp. 278-282, 2009.
[35] A. Setayeshmehr, A. Akbari, H. Borsi and E. Gockenbach, “On-line monitoring and diagnoses
of power transformer bushings”, IEEE Trans. Dielectr. Electr. Insul., Vol. 13, No. 3, pp. 608-
615, 2006.
[36] C. Gonzalez, J. Pleite, V. Valdivia and J. Sanz, “An overview of the on Line application of
Frequency Response Analysis (FRA)”, IEEE Int’l. Sympos. Indust. Electronics (ISIE), pp. 1294-
1299, 2007.
[37] N. Pinhas, S. Islam and J. Hullett, “On the development of transfer function method for fault
identification in large power transformers on load”, IEEE Conf. Electr. Insul. Dielec.
Phenomena, Vol.2, pp. 747-751, 2000.
[38] K. G. N. B. Abeywickrama, Y. V. Serdyuk and S. M. Gubanski, “Effect of core magnetization on

Frequency Response Analysis (FRA) of power transformers”, IEEE Trans. Power Del., Vol. 23,
No. 3, pp. 1432-1438, 2008.
[39] K. G. N. B. Abeywickrama, A. D. Podoltsev, Y. V. Serdyuk and S. M. Gubanski, “Computation of

parameters of power transformer windings for use in frequency response analysis”, IEEE
Trans. Magn., Vol. 43, No. 5, pp. 1983-1990, 2007.
[40] K. G. N. B. Abeywickrama, Y. V. Serdyuk, and S. M. Gubanski “High-frequency modeling of

power transformers for use in Frequency Response Analysis (FRA)”, IEEE Trans. Power Del.,
Vol. 23, No. 4, pp. 2042–2049, 2008.
[41] S. D. Mitchell and J. S. Welsh, “Modeling power transformers to support the interpretation of
Frequency-Response Analysis”, IEEE Trans. Power Del., Vol. 26, No. 4, pp. 2705-2717, 2011.
[42] A. Shintemirov, W. H. Tang and Q. H. Wu, “A hybrid winding model of disc-type power
transformers for frequency response analysis", IEEE Trans. Power Del., Vol. 24, No. 2, pp.
730-739, 2009.
[43] A. Shintemirov, W. J. Tang, W. H. Tang, Q. H. Wu, “Improved modelling of power transformer

winding using bacterial swarming algorithm and frequency response analysis”, Elect.Power
Sys. Research, Vol. 80, pp. 1111-1120, 2010.
263
Bibliography
[44] A. A. Reykherdt, V. G. Davydov, “Effects of test cable ground extensions on repeatability of

frequency response analysis measurements on power transformers”, IEEE Electr. Insul. Mag.,
Vol. 28, No. 3, pp. 26-31, 2012.
[45] J. A. S. B. Jayasinghe, Z. D. Wang, P. N. Jarman, and A. W. Darwin “Investigations on sensitivity

of FRA technique in diagnostic of transformer winding deformations”, IEEE Int’l. Sympos.
Electr. Insul., Indianapolis, IN, USA, pp. 496-499, 2004.
[46] J. A. S. B. Jayasinghe, Z. D. Wang, P. N. Jarman, and A. W. Darwin “Winding movement in

power transformers: A comparison of FRA measurement connection methods”, IEEE Trans.
Dielectr. Electr. Insul., Vol.13, No. 6, pp. 1342-349, 2006.
[47] Y. Li, G. Liu, L. Zhang, L. Zhang and Z. Lin “Transformer winding deformation diagnosis using
middle band frequency response analysis”, Int’l. Conf. Solid Dielec., Winchester, UK, July 8-13,
pp. 677-680, 2007.
[48] M. Wang, A. J. Vandermaar, and K. D. Srivastava, “Transformer winding movement

monitoring in service—Key factors affecting FRA measurements”, IEEE Electr. Insul. Mag.,
Vol. 20, No. 5, pp. 5-12, 2004.
[49] E. Al Murawwi, R. Mardiana, C. Q. Su, “Effects of terminal connections on sweep frequency

response analysis of transformers”, IEEE Electrical Insulation Magazine, Vol. 28, pp. 8-13,
2012.
[50] E. Al Murawwi, R. Mardiana, C. Q. Su, “Effect of terminal connections on SFRA results of

three-winding power transformers”, IEEE International Conference on Electric Power and
Energy Conversion Systems (EPECS’11), pp. 1-6, 2011.
[51] M. Bagheri, Mohammad S. Naderi, T. Blackburn, B. T. Phung, “Dean-Stark vs FDS and KFT
methods in moisture content recognition of transformers”, IEEE Int’l Conf. Power Energy
(PECON’12), Kota Kinabalu, Malaysia, pp. 712-717, 2012.
[52] M. Bagheri, B. T. Phung, T. Blackburn, A. Naderian “Influence of Temperature on Frequency

Response Analysis of Transformer Winding,” IEEE International Electrical Insulation
Conference (EIC’13), Ottawa, Canada, 2-5 June, pp. 40-44, 2013.
[53] M. Florkowski, B. Florkowska, J. Furgal,P. Pajak, “Impact of oil and temperature on initial
voltage distributions in transformer windings at ultra fast stresses”, IEEE Conf. Electr Insul.
and Dielectr. Phenomena (CEIDP’10), pp. 1-4, 2010.
[54] M. Florkowski, B. Florkowska, J. Furgal,P. Pajak, “Influence of oil temperature on frequency

characteristics of disk and layer transformer windings’, IEEE Conf. Electr Insul. and Dielectr.
Phenomena (CEIDP’11), pp. 203-206, 2011.
[55] A. Abu-Siada, N. Hashemnia, S. Islam, M. A. S. Masoum, “Understanding power transformer

frequency response analysis signatures”, IEEE Electr. Insul. Mag.,Vol. 29, No. 3, pp. 48-56,
2013.
[56] R. Yadav, A. Venkatasami, “Frequency response analysis as a technique for moisture

diagnostics of power transformers”, Int’l Conf. on Condition Monitoring and Diagnosis
(CMD’08), pp. 906-908, 2008.
[57] E. Bjerkan, “High frequency modeling of power transformer”, PhD Thesis, Norwegian
University of Science and Technology (NTNU), 2005.
[58] S.V. Kulkarni and S.A. Khaparde, Transformer Engineering Design and Practice, Marcel
Dekker, Inc., New York , USA, 2004.
264
Bibliography
[59] R. M. Del Vecchio, B. Poulin, P. T. Feghali, D. M. Shah, R. Ahuja. Transformer Design Principles
(Second ed.), CRC Press (Taylor and Francis Group), FL, USA, 2010.
[60] D. Allan, H. Moore, Electric Power Transformer Engineering, CRC Press LLC, FL, USA, 2004.
[61] Available online, http://www.nizkoramnik.ru/en/uslugi/perevozka_transformatorov.
[62] Available online, http://www.cmpco.com/OurCompany/News/2009/news090928.html.
[63] Available online, http://www.sns-international.com.tr.
[64] Available online, http://www.ruslanint.com/GALLERY?ID=25.
[65] IEEE Std. 62-1995; IEEE Guide for diagnostic field testing of electric power apparatus, Part 1:
Oil filled power transformers, regulators, and reactors, 1995.
[66] IEC Standard 60076-5; Power transformer: Ability to withstand short circuit, edition 3.0,
2006.
[67] V. Sokolov and B. Vanin “Experience with detection and identification of winding buckling in
power transformers,” Proc. 68th Annu. Int’l. Conf. Doble Clients, pp. 1-13, 2001.
[68] W. H. Tang, A. Shintemirov, Q. H. Wu, “Detection of minor winding deformation fault in high
frequency range for power transformer.” IEEE Power and Energy Society General Meeting,
2010, pp. 1-6.
[69] J. A. S. B. Jayasinghe, Z. D. Wang, P. N. Jarman,A. W. Darwin “Investigations on sensitivity of

FRA technique in diagnosis of transformer winding deformations,” IEEE Int’l. Sympos. Electr.
Insul. (ISEI’04), 2004, pp. 496-499.
[70] S. M. Islam “Detection of shorted turns and winding movements in large power transformers
using frequency response analysis," IEEE Power Engineering Society, Winter Meeting, vol. 3,
2000, pp. 2233-2238.
[71] M. Wang, A. J. Vandermaar,K. D. Srivastava “Transformer winding movement monitoring in

service - key factors affecting FRA measurements,” IEEE Electr. Insul. Mag., vol. 20, 2004, pp.
5-12.
[72] M. Bagheri, Mohammad S. Naderi, T. Blackburn, D. Zhang “Transformer frequency response

analysis: A mathematical approach to interpret mid-frequency oscillations,” IEEE Int’l Conf.
Power Energy (PECON’12), Kota Kinabalu, Malaysia, 2-5 Dec., 2012, pp. 927-931.
[73] M. Wang, A. J. Vandermaar,K. D. Srivastava “Improved detection of power transformer

winding movement by extending the FRA high frequency range,” IEEE Trans. Power Del., vol.
20, 2005, pp. 1930-1938.
[74] Frequency response analysis on winding deformation of power transformers, People’s

Republic of China, Electric Power Industry Standard, DL/T911-2004, ICS27.100, F24,
Document No. 15182-2005, June 1st, 2005.
[75] R. Wimmer , S. Tenbohlen , M. Heindl , A. Kraetge , M. Kruger and J. Christian “Development of

algorithms to assess the FRA,” 15th Int. Symp. High Voltage Engineering, Paper No. T7-523,
Ljubljana, Slovenia, 2007.
[76] P. M. Joshi and S. V. Kulkarni, "Transformer winding diagnostics using deformation

coefficient", IEEE Power and Energy Soc., General Meeting, Conversion and Delivery of
Electrical Energy in the 21st Century, pp. 1-4, 2008.
265
Bibliography
[77] P.M. Joshi, S.V. Kulkarni. “Diagnostic method for determining deformations in a transformer
winding”, United States Patent Application, Publication No. US 2010/0211339 A1; Aug 2010.
[78] J. C. Maxwell, A Treatise on Electricity and Magnetism. Vol. 2, 3rd Ed., Oxford, Clarendon
Press, Section 696-701, p. 334, p. 336, p. 339, 1892.
[79] N. Rayleigh, “Inductance of a short cylindrical current sheet”, Proc. Roy. Soc., Vol. 32, pp.104-
141, 1881.
[80] T. R. Lyle, “On circular filaments or circular magnetic shells equivalent to circular coils, and
on the equivalent radius of a coil”, Phil. Mag., Vol. 3, p. 310, 1902.
[81] S. Butterworth, “On the coefficients of mutual induction of eccentric coils”, Phil. Mag., ser. 6,
Vol. 31, pp. 443–454, 1916.
[82] C. Snow, Formulas for Computing Capacitance and Inductance, National Bureau of Standards
Circular 544. Washington DC, Dec. 1954.
[83] E. B. Rosa, “Calculation of the self-inductance of single-layer coils”, Bureau of Standards

Bulletin, Vol. 2, pp. 161-187, 1906.
[84] H. L. Curtise and M. Sparks, “Formulas, tables, and curves for computing the mutual
inductance of two coaxial circles”, Bureau of Standards Bulletin. Vol. 19. Scientific paper 492,
pp. 541-576, 1924.
[85] F. W. Grover, Inductance Calculations: Working Formulas and Tables, New York: Dover,
Power Publisher Inc., 1964.
[86] S. Babic, F. Sirois, C. Akyel, C. Girardi, “Mutual inductance calculation between circular
filaments arbitrarily positioned in space: Alternative to Grover’s formula”, IEEE Trans.Magn,
vol. 46, pp. 3591-3600, 2010.
[87] J. T. Conway, “Inductance calculations for non-coaxial coils using Bessel functions,” IEEE
Trans. Magn., vol. 43, no. 3, pp. 1023–1034, Mar. 2007.
[88] K. A. Wirgau, “Inductance calculation of an air-core disk winding”, IEEE Trans. Power App.
Syst., Vol. 95, pp. 394-400, 1976.
[89] C. Alvarez-Marino, F. de Leon, X. M. Lopez-Fernandez, “Equivalent circuit for the leakage

inductance of multiwinding transformers: Unification of terminal and duality models”, IEEE
Trans. Power Del.,Vol. 27, 2012, pp. 353-361.
[90] G. Kirchhoff, Gesammelte Abhandlungen, Published by Leipzig: J.A. Barth, p. 177, 1882.
[91] H. Nagaoka, “The inductance coefficient of solenoids”, Vol 27, Jour. Coll. Sci. Tokyo, article 6,
1909.
[92] K. Karsai, D. Kerenyi, Large Power Transformer, ELSEVIER, New York, pp.187–214, 1987.
[93] Mohammad S. Naderi, M. Vakilian, T.R. Blackburn, B.T. Phung, Mehdi S. Naderi and A. Nasiri,
“A hybrid transformer model for determination of partial discharge location in transformer
winding”, IEEE Trans. Dielectr. Electr. Insul., Vol. 14, No. 2, pp.436-443, 2007.
[94] R. Rudenberg, Electrical Shock Waves in Power Systems: Traveling Waves in Lumped and
Distributed Circuit Elements, Harvard University Press. Cambridge, Massachusetts, pp. 171-
174, 1968.
266
Bibliography
[95] L. V. Bewley, Traveling Waves on Transmission Systems, Dover Publication, Inc..New York,
pp. 421-423, 1963.
[96] A. Shintemirov and Q.H. Wu “Transfer function of transformer winding for frequency
response analysis based on traveling wave theory”, Proc. Int’l. Control Conf. (ICC), Glasgow,
Scotland, UK, p. 6, 2006.
[97] M. Bagheri, Mohammad S. Naderi, T. Blackburn and T. Phung, “Case study on FRA capability
in detection of mechanical defects within a 400MVA transformer,” CIGRE, Paris, France, pp.
1-9, 2012.
[98] A. Akbari, P. Werle, H. Borsi and E. Gockenbach, “Transfer function-based partial discharge
localization in power transformers: a feasibility study,” IEEE Trans. Electr. Insul. Mag., Vol.
18, No. 5, pp. 22-32, 2002.
[99] M. Bagheri, A. Hekmati, R. Heidarzadeh and M. S. Naderi, “Impulse voltage distribution in

intershield disk winding vs interleaved and continuous disk winding in power transformer,”
IEEE Power and Energy Conf. (PECON), pp 387-392. 2008.
[100] M. Bagheri, M. Vakilian, A. Hekmati and R. Heidarzadeh, “Influence of electrostatic shielding

of disc winding on increasing the series capacitance in transformer”, IEEE PowerTech,
Lausanne, Switzerland, pp.1780-1784, 2007.
[101] J. Stefan, Circular coil induction coefficient calculation, Ann. Phys., pp.112–113, 1884.
[102] E. Rahimpour, J. Christian, K. Feser,H. Mohseni, "Transfer function method to diagnose axial
displacement and radial deformation of transformer windings", IEEE Trans. Power Del., vol.
18, pp. 493-505, 2003.
[103] S. Chattopadhyay, “Buckling behavior of a superconducting magnet coil”, Int’l Jour. Solids and
Structures, Vol. 43, 2006, pp. 5158-5167.
[104] C. R. Paul, Inductance Loops and Partial, Published by John Wiley, Ed. 1, 2009.
[105] Megger, “A Stitch in Time: The Complete Guide to Electrical Insulation Testing”, 2006,
Available online, http://www.biddlemegger.com/biddle/Stitch-new.pdf.
[106] D. F. Garcia, B. Garcia, J. C. Burgos, “A review of moisture diffusion coefficients in transformer

solid insulation-part 1: coefficients for paper and pressboard”, IEEE Electr. Insul. Mag. Vol.
29, No. 1, pp. 46-54, 2013.
[107] Warren L. McCabe, Julian C. Smith and P. Harriott, Unit operations of chemical engineering,
McGraw-Hill, 7th ed., Ch. 4, 2005.
[108] W. Guidi and H. Fullerton “Mathematical methods for prediction of moisture take-up and
removal in large power transformers”, Proceedings of IEEE Winter Power Meeting, no. C-74,
pp. 242–244, 1974.
[109] Y. Du, A. V. Mamishev, B. C. Lesieutre, M. Zahn, “Measurement of moisture diffusion as a

function of temperature and moisture concentration in transformer pressboard”, IEEE Conf.
Electr. Insul. Dielectr. Phenomena, Annual Report, Vol. 1, pp. 341-344, 1998.
[110] Y. Du, M. Zahn, B. C. Lesieutre, A. V. Mamishev,S. R. Lindgren, “Moisture equilibrium in

transformer paper-oil systems”, IEEE Electr. Insul. Mag., Vol. 15, No. 1, pp. 11-20, 1999.
[111] L. Jian, Z. Zhaotao, S. Grzybowski, L. Yu, “Characteristics of moisture diffusion in vegetable

oil-paper insulation”, IEEE Trans. Dielectr. Electr. Insul., Vol. 19, No. 5, pp. 1650-1656, 2012.
267
Bibliography
[112] B. Garcia, J. C. Burgos, A. M. Alonso,J. Sanz, “A moisture-in-oil model for power transformer
monitoring - Part I: Theoretical foundation”, IEEE Trans. Power Del., Vol. 20, No. 2, pp. 1417-
1422, 2005.
[113] T. V. Oommen, “Moisture equilibrium in paper-oil insulation systems”, in Proc. of 16th

Electrical/Electronics Insulation Conf., Chicago, IL, pp. 162-166, 1983.
[114] J. Fabre and A. Pichon, “Deteriorating processes and products of paper in oil. Application to
transformers,” in Proc. Int’l Conf. Large High Voltage Electric System, Paris, France, paper
137, 1969.
[115] P. J. Griffin, “Water in transformers-so what!”, in National Grid Condition Monitoring Conf.,
pp. 1-16, May 1996.
[116] E. J. Murphy, S. O. Morgan, “The dielectric properties of insulating materials, Part III:
Alternating and direct current conductivity”, BELL System Technical Journal, no. 18, 502-
537, 1939.
[117] N. Abeywickrama, C. Ekanayake, Y. V. Serdyuk and S. M. Gubanski, “Effects of the insulation

quality on the frequency response of power transformers”, Journal of Electrical Engineeering
and Technology, Vol. 1, no. 4, pp. 534-542, 2006.
[118] S. M. Gubanski, P. Boss, G. Csepes, V. Der Houhanessian, J. Filippini, P. Guuinic, U. Gafvert, V.

Karius, J. Lapworth, G. Urbani, P. Werelius, W. Zaengl, Dielectric Response Methods for
Diagnostics of Power Transformers, CIGRE Technical Brochure, No. 254, Paris 2004.
[119] C. Ekanayake, Application of Dielectric Spectroscopy for Estimating Moisture Content in

Power Transformers, Lic. Thesis, Chalmers University of Technology (CTH), Gothenburg,
Sweden, 2003, Techn. Rep. No. 465L.
[120] B. García, J. C. Burgos and A.M. Alonso, “Winding deformations detection in power
transformers by tank vibrations monitoring”, Electric Power Syst. Research, Vol. 74, pp. 129-
138, 2004.
[121] B. García, J.C. Burgos and A.M. Alonso, “Transformer tank vibration modeling as a method of
detecting winding deformations-part I: theoretical foundation”, IEEE Trans. Power Del., Vol.
21, pp. 157-163, 2006.
[122] B. García, J.C. Burgos and A.M. Alonso, “Transformer tank vibration modeling as a method of
detecting winding deformations-part II: experimental verification”, IEEE Trans. Power Del.,
Vol. 21, pp. 164-169, 2006.
[123] T. He, J-Di. Wang, J. Guo, H. Huang, X.-X.Chen and J. Pan, “A vibration based condition
monitoring system for power transformers”, Asia-Pacific Power and Energy Eng. Conf.
(APPEEC), pp. 1-4, 2009.
[124] M.A. Hejazi, M. Choopani, M. Dabir and G.B. Gharehpetian, “Effect of antenna position on on-
line monitoring of transformer winding axial displacement using electromagnetic waves”,
IEEE 2nd Int'l. Power Energy Conf. (PECon), pp. 44-49, 2008.
[125] M. Akhavanhejazi, G.B. Gharehpetian, R. Faraji-Dana, G.R. Moradi, M. Mohammadi and H.A.
Alehoseini, “A new on-line monitoring method of transformer winding axial displacement
based on measurement of scattering parameters and decision tree”, Expert Syst.
Applications, Vol. 38, pp. 8886-8893, 2011.
[126] M. A. Hejazi, G. B. Gharehpetian, G. Moradi, H. A. Alehosseini and M. Mohammadi, “Online

monitoring of transformer winding axial displacement and its extent using scattering
268
Bibliography
parameters and k-nearest neighbour method”, IET Proc. Generation, Transmission

Distribution, Vol. 5, pp. 824-832, 2011.
[127] P. M. Joshi and S. V. Kulkarni, “A novel approach for online deformation diagnostics of
transformer windings”, IEEE Power and Energy Society, General Meeting, pp. 1-6, 2010.
[128] N. Shu, C. Zhou, F. Hu, Q. Liu and L. Zheng, “Study on ultrasonic measurement device for
transformer winding deformation”, IEEE Int’l. Conf. Power Syst. Techn., (POWERCON), Vol.3,
pp. 1401-1404, 2002.
[129] M. Alpatov, “On-line detection of winding deformation”, IEEE Int’l. Sympos. Electr. Insul., pp.
113-116, 2004.
[130] S.-b. Gao and G. Wang, “Study on on-line monitoring of windings deformation of power
transformer”, IEE 8th Int’l. Conf. Developments in Power System Protection, pp. 335-338
Vol.1, 2004.
[131] G. Hu, L. Zhang, X. Wu, Correia, D. and W. He, “Detecting the capacity of distribution
transformer based on an on-line method”, Asia-Pacific Power and Energy Eng. Conf.
(APPEEC), pp. 1-4, 2011.
[132] A. Palani, S. Santhi, S. Gopalakrishna and V. Jayashankar, “Real-time techniques to measure

winding displacement in transformers during short-circuit tests”, IEEE Trans. Power Del.,
Vol. 23, pp. 726-732, 2008.
[133] D. K. Xu and Y. M. Li, “A simulating research on monitoring of winding deformation of power

transformer by on-line measurement of short-circuit reactance”, IEEE Int’l. Conf. Power
System Technology (POWERCON), Vol.1, pp. 167-171, 1998.
[134] Z. Berler, P. Butyrin and M. Alpatov,”On-line diagnostics of mechanical deformations of

winding”, Proceedings of TechCon’05, pp. 1-4, 2005.
[135] E. Arri, A. Carta, F. Mocci and M. Tosi, “Diagnosis of the state of power transformer windings
by on-line measurement of stray reactance”, IEEE Trans. Instrum. Meas., Vol. 42, pp. 372-
378, 1993.
[136] P. Li, B. G. Zhang, Z.G. Hao, X.J. Hu and Y.L. Chu, “Research on monitoring of winding
deformation of power transformer by on-line parameter estimation about leakage
inductance”, IEEE Int’l. Conf. Power Syst. Techn. (POWERCON), pp. 1-6, 2006.
[137] Z. G. Hao, B. H. Zhang, C. G. Yan, B. R. Shao, X. F. Bo and Z. Q., “Research on integration of

transformer protection and winding deformation detecting”, IEEE Int’l. Conf. Power Syst.
Techn. (POWERCON), pp. 1-8, 2010.
[138] A. Setayeshmehr, H. Borsi and E. Gockenbach, “Multi Channel On-Line Monitoring System for
Power Transformers”, Int’l. ETG-Kongress Düsseldorf, Germany, pp. 369-374, 2009.
[139] A. Setayeshmehr, I. Fofana, A. Akbari, H. Borsi and E. Gockenbach “Effect of temperature,

water content and aging on the dielectric response of oil-impregnated paper”, IEEE Int’l.
Conf. Dielectr. Liquids (ICDL), pp. 1-4, 2008.
[140] W. J. McNutt, W. M. Johnson, R. A. Nelson and R. E. Ayers, “Power Transformer Short-Circuit

Strength-Requirements, Design, and Demonstration”, IEEE Trans. Power App. Syst. Vol. 89,
pp. 1955-1969, 1970.
[141] Trafo Union Co.,”Transformer design documents”, TUB 27000.01.0 and 27100.10.0.
Germany.
269
Bibliography
[142] M. Florkowski, “Exploitation stresses and challenges in diagnostics of electrical industrial

equipment”, IEEE Int’l. Sympos. Industrial Electronics (ISIE), pp. 15-25, 2011.
[143] M. Florkowski and J. Furgal, “Identification of winding faults in electric machines using a high
frequency method”, IEEE Int’l. Conf. Electric Machines & Drives (IEMDC) ., pp. 590-595, 2007.
[144] M. Wang, A. J. Vandermaar and K. D. Srivastava, “Improved detection of power transformer

winding movement by extending the FRA high frequency range”, IEEE Trans. Power Del., Vol.
20, pp. 1930-1938, 2005.
[145] K. Feser, J. Christian, C. Neumann, U. Sundermann, T. Leibfried, A.Kachler, and M. Loppacher,

“The transfer function method for detection of winding displacements on power
transformers after transport, short circuit, or 30 years of service”, CIGRE, Paris, France,
paper no.12/33-04, 2000.
[146] B. Barkat and B. Boashash, “A high-resolution quadratic time-frequency distribution for

multicomponent signals analysis”, IEEE Trans. Signal Processing, Vol. 49, pp. 2232-2239,
2001.
[147] N. Abeywickrama, D. Bormann, J.-L. Bermudes, O. Kouzmine and P. Werle, “Sensitivity of FRA
in detecting transformer faults: Results of a measurement campaign”, in CIGRE Kyoto Japan,
2011.
[148] A. Wilk and D. Adamczyk, “Investigations on sensitivity of FRA method in diagnosis of

interturn faults in transformer winding”, IEEE Int’l. Sympos. Industrial Electronics (ISIE), pp.
631-636, 2011.
[149] J. Bak-Jensen, B. Bak-Jensen,S. D. Mikkelsen, “Detection of faults and ageing phenomena in

transformers by transfer functions”, IEEE Trans. Power Del., vol. 10, no. 1, pp. 308-314, 1995.
[150] H. B. Dwight, Tables on Integrals and Other Mathematical Data, Macmillan, New York, 1947.
[151] R. C. Degeneff, J. C. Crouse, “Electrostatic shielding of nonsequential disc windings in

transformers”. U.S Patent 4243966, Jan 1981.
[152] R.M. Del Vecchio, B. Poulin, R. Ahuja, “Calculation and measurement of winding disk
capacitances with wound-in-shields”, IEEE Trans. Power Del., vol. 13, No. 2, pp. 503-509,
April 1998.
[153] M. Bélanger, “Applying an online dry-out process to power transformers”, Electricity Today
Magazine, Issue 5, pp. 50-54, 2003.
[154] W. Knapek, J. Foley, “Dielectric response analysis of transformers,” Neta World Journal
(Feature), Winter 2009-2010.
[155] S. M. Gubanski, J. Blennow, L. Karlsson, K. Feser, S. Tenbohlen, C. Neumann, H. Moscicka-

Grzesiak, A. Filipowski, L. Tatarski, “Reliable diagnostics of HV transformer insulation for
safety assurance of power transmission system,” CIGRE, PARIS, pp. 1-8, 2006.
[156] IEC Std. 60814, “Insulating liquid oil- impregnated paper and press board determination of
water by coulometric Karl Fischer titration,” 1997-8.
[157] Sample Collection Procedures for Insulating Fluid Samples, An online report by TJ/H2B.
[158] C. Aurand, Moisture Determination by Karl Fischer Titration, Background of the Chemistry
and Recent Developments Supelco/Sigma-Aldrich, Presentation in Bellefonte, Pa, 2010.
270
Bibliography
[159] Technical brochure on coulometer for Karl Fischer titrations, Metrohm AG, Ionenstrasse,
Switzerland.
[160] J. Fabre and A. Pichon, “Deteriorating process and products of paper in oil application to
transformers”, International Conference on Large High Voltage Electric System, CIGRE, Paris,
France, Paper 137, 1960.
[161] T. V. Oommen, “Moisture equilibrium for paper-oil system”, Proceeding of

Electrical/Electronic of Insulation Conference, Chicago, IL, pp. 162-166, Oct. 1983.
[162] E. T. Norris, “High voltage power transformer insulation”, Proceeding I.E.E, Vol. 110, no. 2,
pp. 428-440, February 1994.
[163] P. J. Griffin, C. M. Bruce, J.D. Christie, “Comparison of water equilibrium in silicone and
mineral oil transformes”, Annual Interbational Conference of Doble Clients, 1988.
[164] Relative Humidity Technology Vaisala HUMICAP for measuring relative humidity”, Ref.
B210781EN-B, 2009, Available online, http://www.ni.com/white-paper/9887/en.
[165] A. Tutkevick, “Power transformer assembly”, Union of Soviet Socialist Republics Press, 1975.
[166] M. Koch, M Kruger, S. Tenbohlen, “Comparing various moisture determination methods for
power transformers”, CIGRE , 6th Southern Africa Regional Conference, pp. 1-6, 2009.
[167] A. Bouaïchz, I. Fofana, M. Farzaneh, A. Setayeshmehr, H. Borsi, E. Gockenbach, A. Beroual,

Ngnui Thomas Aka, “Dielectric spectroscopy techniques as quality control tool: A feasibility
study,” IEEE Electr. Insul. Mag., vol. 25, issue 1, pp. 6-14, 2009.
271

Kfkgig Fifd 4u

Uploaded by

Copyright:

Available Formats

You might also like

Kfkgig Fifd 4u

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kfkgig Fifd 4u

Uploaded by

Copyright:

Available Formats

Transformer Winding Deformation and Insulation

Characteristic Effects on Frequency Response Analysis

Downloaded from http://hdl.handle.net/1959.4/53515 in https://

Supervisor: Dr.Toan PHUNG

A THESIS IN FULFILMENT OF THE REQUIREMENT FOR

SCHOOL OF ELECTRICAL ENGINEERING AND TELECOMMUNICATIONS

First name: Mehdi Other name/s:

Abbreviation for degree as given in the University calendar: Ph.D.

School: Electrical Engineering and Telecommunications Faculty: Engineering

Abstract 350 words maximum:

……………………………………………………… …………………………………………… ……14/06/2014……….

FOR OFFICE USE ONLY Date of completion of requirements for Award:

IN LOVING MEMORY OF MY MOTHER

I would also like to give my special appreciation to my co-supervisor, A/Prof. Trevor

A Sinusoidal signal amplitude Ch.2

Table 2.1 FRA measurement connections (phases A, B and C) 30

1.1 Problem Statement

1.3 Current State of the Art

1.4 Research Objectives

1.5 Thesis Overview

Chapter 7 experimentally discusses the effect of temperature and moisture content

1.6 Key Contributions

 Demonstrated analytically how axial and radial deformation in transformer

1. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, “Advanced

2. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “Transformer Frequency

3. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, Toan Phung

4. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “Influence of Temperature and

6. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung, “Practical

7. Mehdi Bagheri, Mohammad S. Naderi, “Moisture Diagnostics of Power

9. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung, “Online

Transformer Winding”, IEEE International Conference on Power and Energy

11. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, Damin Zhang,

13. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung,

16. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung

17. Mehdi Bagheri, Mohammad S. Naderi, Trevor Blackburn, B.T. Phung

18. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “On-line Transformer

20. Mehdi Bagheri, B. T. Phung, Trevor Blackburn “Impact of Transformer

21. Mehdi Bagheri, B. T. Phung, Trevor Blackburn, “Impacts of Transformer

24. Mehdi Bagheri, Toan Phung, Trevor Blackburn, “A Practical Solution to

25. Mehdi Bagheri, B. T. Phung, Trevor Blackburn, “The Influence of Dielectric

Transformer Winding Deformation

This Chapter discusses transformer winding deformation and available diagnosis

2.2 Transformer Winding Deformation and Displacement

Nspacer   Dave , (2.1)

2.2.1 Short Circuit Current

Of all the possible causes of transformer failure, mechanical deformation of windings as a

2.2.1.1 Radial Forces

2.2.1.2 Axial Forces

2.2.2 Transformer Transportation Causing Active Part Displacement

Transformer transportation is one of the major causes in transformer active part

In each region, based on available transportation infrastructures one of the mentioned

Figure 2.3. Transformer transportation schematic.

 Fradial   Fwind   Fpredicted   mg   Fstop

2.3 Winding Deformation Diagnosis Methods

2.3.1 Short Circuit Impedance Measurement

SCI measurements can be performed on single or three-phase transformers, usually on the

Figure 2.6. Schematic of transformer core and windings.

2.3.2 Transfer Function Methods (FRA/LVI)